Preventable Drug-Related Hospital Admissions

Pharmacoepidemiology

Preventable Drug-Related Hospital Admissions Almut G Winterstein, Brian C Sauer, Charles D Hepler, and Charles Poole

OBJECTIVE:

To estimate the prevalence of preventable drug-related hospital admissions (PDRAs) and to explore if selected study characteristics affect prevalence estimates. METHODS:

Keyword search of MEDLINE (1966–December 1999), International Pharmaceutical Abstracts (1970–December 1999), and hand search. Two reviewers independently selected studies published in peer-reviewed journals and extracted crude prevalence estimates and study characteristics. Trials had to specifically address consequences of drug therapy requiring hospital admission and include a quantitative preventability assessment. Stratified analysis and meta-regression were used to explore the association between study characteristics and prevalence estimates.

DATA SYNTHESIS:

Fifteen studies reported a median PDRA prevalence of 4.3% (interquartile range [IQR] 3.1–9.5%). The median preventability rate of drug-related admissions was 59% (IQR 50–73%). No evidence of publication bias related to study size could be determined. Because the individual study results were highly heterogeneous (Cochran’s Q = 176, df = 14; p < 0.001), no metaanalytic summary estimate was computed. Stratified analysis suggested an association between prevalence estimates and 3 study characteristics: exclusion of first admissions (readmission studies: average PDRA prevalence of 14.0 %, estimated prevalence OR = 3.7); mean age of admissions >70 (OR = 2.1); and inclusion of “indirect” drug-related morbidity, such as omission errors or therapeutic failure (OR = 1.9). There was little evidence of other associations with prevalence estimates, such as selection of specific hospital units, exclusion/inclusion of planned admissions, country, and specified methods of PDRA case ascertainment.

CONCLUSIONS:

Drug-related morbidity is a significant healthcare problem, and a great proportion is preventable. Study methods in prevalence reports vary and should be considered when interpreting findings or planning future research.

KEY WORDS: adverse drug events, adverse effects, meta-analysis.

Ann Pharmacother 2002;36:1238-48.

recent report of the Institute of Medicine (IOM)1 reviewed the prevalence and significance of human error in health care and its implication for patient safety. In the report, the authors concluded that medication-related errors are “one of the most common types of errors…. [S]ubstantial numbers of individuals are affected, and it accounts for a sizeable increase in health care costs.”

A

Author information provided at the end of the text. Preliminary results of this study have been presented as posters at the 36th Annual Drug Information Association Meeting, San Diego, June 2000, and 16th International Conference on Pharmacoepidemiology, ISPE, Barcelona, Spain, August 16-20, 2000; and as an oral presentation at the American Pharmaceutical Association Annual Meeting in San Francisco, March 18-21, 2001.

1238

■

The Annals of Pharmacotherapy

■

As a consequence, public awareness about patient safety has increased, but information about the prevalence of drug-related morbidity (DRM) and the proportion preventable is widely scattered, and its generalizability is still unknown. Prevalence estimates that focus on medical error and preventability, including the findings of the IOM report, are controversial and continue to be challenged. Major limitations that have been discussed in the literature2-6 concern unrepresentative samples, unreliable or untested methods for case finding and preventability assessment, subjective judgment, and publication biases. Little generalizable information exists about the incidence of DRM in ambulatory care. The IOM report1 quoted incidence rates of adverse drug events (ADEs) occurring in acute care that were based on 2 large-scale multisite

2002 July/August, Volume 36

Downloaded from aop.sagepub.com at UNIV OF UTAH SALT LAKE CITY on August 3, 2016

www.theannals.com

studies representative of 2 states. For ambulatory care, it suggested, “There is evidence indicating that ADEs account for a sizable number of admissions to inpatient facilities, but we do not know what proportion of these ADErelated admissions are attributable to errors.” Because of the controversy, proposed limitations, and lack of generalizability, there is a need to formally analyze studies on DRM associated with ambulatory care and to synthesize findings into reliable prevalence estimates. This systematic review was designed to address the following research questions: What is the reported prevalence of preventable drug-related hospital admissions (PDRAs)? Do selected study characteristics affect these prevalence estimates, thus compromising their generalizability? Methods For the purpose of this review, we have used DRM as a general term for drug-related adverse patient outcomes. DRM includes outcomes by an adverse or toxic effect (direct) or failure to obtain the necessary result within a reasonable time, including occasions where an indicated therapy was either not used or was pharmaceutically ineffective (indirect). LITERATURE SEARCH

A search of MEDLINE (1966–December 1999) and International Pharmaceutical Abstracts, (IPA) (1970–1999) electronic databases was conducted with the following key word strategy: (ADR or ADRs or adverse drug reaction or adverse reaction or adverse event or drug-related or drug induced or iatrogenic) and (preventable or preventability or avoidable). MEDLINE revealed 209 publications and IPA 242. Additional information was found by following the reference citations from retrieved articles and by hand search. Two reviewers independently selected studies according to the following explicit criteria. SELECTION CRITERIA

Peer-reviewed trials in any language conducted in any medical inpatient service from general or teaching hospitals in any country were included. Studies had to specifically address consequences of drug therapy that required hospital admission. They had to assess both the causal relationship between pharmacotherapy and patient morbidity, and preventability, along with the necessary information to calculate prevalences. Studies were excluded if case identification was based only on screening instruments, computerized alert systems, or spontaneous reports of unknown validity. The trials had to take a comprehensive approach to drug therapy as a cause of patient injury. We excluded studies that limited their scope to specific drug treatments (e.g., theophylline toxicity), indications (e.g., diabetes, asthma), injuries (e.g., hypoglycemia), or only 1 drug-therapy problem (e.g., drug– drug interactions, adverse reactions). www.theannals.com

DATA EXTRACTION

We developed a data collection template prior to our review to increase reliability and efficiency. Two reviewers used the template independently to extract data from each report. When differences occurred, the reviewers discussed the disparity and came to a consensus. In 1 case, the authors of a study were contacted for clarification. The following information was extracted from each included study (Table 1):7-21 total sample size, period, and sample descriptors; the number of drug-related admissions (DRAs); and the number of PDRAs.5 In some studies, the investigators assigned confidence descriptions to their assessments of causality, the contribution to hospital admission (e.g., major or contributing reason for admission), or preventability (e.g., definitely or probably preventable). Cases were included as drug-related and preventable if the investigators had considered them. In every case, these had been labeled as definite, probable, or possible. Cases labeled doubtful or not assessable were excluded. We also extracted reported types of drug-therapy problems, e.g., overdose, lack of drug, wrong drug, and methods of case ascertainment such as case search and assessment methods, data sources used for these assessments, the number of review panelists, and how consensus was expressed (Table 2).7-25 The reviewed studies used different judgment methods for the assessment of causality and preventability. We classified the applied judgment methods according to the system of Ashton et al.26 (Table 2). Three approaches are defined as implicit, explicit, and structured implicit review. In an implicit approach, the reviewer would compare a certain circumstance with his or her own knowledge, opinions, and beliefs. An explicit review provides detailed criteria (e.g., algorithm) that define the elements of the judgment. In a structured implicit approach, the reviewers apply internalized standards but are directed to base their judgment on specific issues and information. When available, explicit criteria for preventability assessments reported by the respective authors are described in Appendix I11-13,19; for causality assessment, the original references are quoted in Table 2. DATA ANALYSIS

The prevalence of DRAs was calculated by dividing the reported number of DRAs by the total number of admissions. The prevalence of PDRAs was obtained by dividing the number of PDRAs by the total number of admissions. The preventability rate was calculated by dividing the number of PDRAs by the total number of DRAs. Prevalence estimates were summarized with the median and interquartile range (IQR) to illustrate the distribution of study findings without consideration of heterogeneity. For statistical analyses, PDRA prevalence estimates were transformed to logits (log odds) to take advantage of the normalizing and variance-stabilizing properties of the logit transformation. Results were transformed back to the original prevalence scale for ease of interpretation. We tested whether the admission studies fulfilled specific requirements


■



■

1239

AG Winterstein et al.

Table 1. Studies Reporting DRAs and PDRAs: Sample Characteristics and Prevalence Estimates Reference

Country

Study Y/Period

Mean Pt. Age (y)

Prevalence n (%) DRA PDRA

Inclusion (+)/Exclusion (–) Criteria

Preventability of DRA n (%)

Darchy et al. (1999)7

France

1994, 12 mo

69

+ all admissions and transfers

41/623 (6.6)

30/623 (4.8)

30/41 (73.2)

Ng et al. (1999)8

Australia

1996, 3 wk

75

+ unplanned admissions – elective admissions and transfers

31/172 (18.0)b

10/172 (5.8)

10/32 (31.3)a

Raschetti et al. (1997)9

Italy

1994–1995, 12 wk

69

+ all admissions from ED

45/1833 (2.5)

25/1833 (1.4)

25/45 (55.6)

Cunningham et UK al. (1997)10

4 wk, each unit

77

+ admissions to all hospitals in Tayside, patients ≥65 y

54/1011 (5.3)b

43/1011 (4.3)

43/54 (79.6)a

Nelson and Tal- US bert (1996)11

1993, 1 mo

47


73/450 (16.2)b

43/450 (9.5)

43/73 (58.9)a

Courtman and Stallings Canada (1995)12

1992–1993, 5 mo

78

+ all admissions, patients ≥65 y; – transfers

21/150 (14.0)b

18/150 (12.0)

18/21 (85.7)a

Dartnell et al. (1996)13

Australia

1994, 1 mo

58 (median)

+ all admissions from ED, hospital stay >24 h

55/965 (5.7)b

36/965 (3.7)

36/55 (65.5)a

Hallas et al. (1992)14

Denmark

1988–1989

68.5

+ ~300 admissions/hospital unit; – planned admissions of non-Odense citizens

8.0b adjusted (n = 1999)

3.8 adjusted

67/143 (46.9)a

UK

10 wk

76

+ admissions (emergency and planned), ≥65 y; – transfers and readmissions

26/416 (6.3)b

13/416 (3.1)

13/26 (50.0)

Nikolaus et al. Germany (1992)16

1987–1990, 36 mo

81

+ readmissions within 3 wk after discharge

22/87 (25.3)

11/87 (12.6)

11/22 (50.0)

Bero et al. (1991)17

US

NR

75

45/224 (20.1)

34/224 (15.2)

34/45 (75.6)

Bigby et al. (1987)18,c

US

73/686 (10.6)

43/686 (6.3)

43/73 (58.9)a

35/834 (4.2)b

19/834 (2.3)

19/35 (54.3)

Lindley et al. (1992)15

+ 1st readmissions within 6 mo, ≥65 y, Medicare, >3 drugs; – admission from nursing home or hospice, no phone, not English-speaking, mentally impaired, no proxy 1983–1984, 24 mo

60

Lakshmanan et US al. (1986)19

1984, 2 mo

54

Trunet et al. (1986)20

France

1978–1981, 33 mo

NA


97/1651 (5.9)

43/1651 (2.6)

43/97 (44.3)

Trunet et al. (1980)21

France

1978–1979, 12 mo

NA

+ all admissions and transfers every other wk

23/325 (7.1)b

14/325 (4.3)

14/23 (60.9)

+ unplanned admissions, ≥1 visit to hospital primary care clinic in last 2 y + all admissions

DRA = drug-related admission; ED = emergency department; NA = not applicable; NR = not reported; PDRA = preventable drug-related admission. a Preventability qualifiers: refs. 10, 11, 13, and 14: definite or possible refs. 8 and 18: definite or probable ref. 12: definite b Causality qualifiers: refs. 11 and 14: definite or probable; dominant, or partly contributing to admission refs. 13 and 21: definite, probable, possible ref. 8: probable and possible; potential contributor to admission refs. 10 and 19: definite or probable ref. 12: major reason for admission ref. 15: possibly or definitely related to admission c Divided causes for hospital admissions into iatrogenic, lack of patient compliance, and natural history of disease. We assumed that lack of patient compliance refers only to pharmacotherapy and included all cases of noncompliance in our analysis. The category iatrogenic encompassed 4 subcategories: adverse drug reactions, failure to follow up, procedure complications, and misdiagnosis. These were not mutually exclusive. From these subcategories, only adverse drug reactions was included because it implies a clear relationship with pharmacotherapy. However, the authors gave some examples for the failure to follow up subcategory that are also drug-related (e.g., pulmonary edema after discontinuation of furosemide).

1240

■


■



www.theannals.com

Preventable Drug-Related Hospital Admissions

for computing a meta-analytic summary estimate. This included the examination for publication bias and for heterogeneity of PDRA prevalence estimates. Heterogeneity was assessed by computing p values for Cochran’s Q statistic.27 Evidence for publication bias, in the form of an association between the logits of published prevalence estimates and sample size, was assessed in 2 ways. First, a funnel plot was examined for asymmetry in the relation between the logits and their precision. The precision of each estimated logit was estimated as the inverse of its estimated variance, computed as the sum of the reciprocals of the numbers of cases and noncases.28 Second, the rank order correlation test of Begg and Mazumdar29 was used to quantify the association between the prevalence estimates and their standard errors. To answer the second research question about possible causes for heterogeneity, we compared prevalence estimates between studies with different characteristics using random-effects meta-regression.30-32 The following characteristics were considered, based on discussions about the

validity of prevalence reports in the literature (Table 3): inclusion versus exclusion of first hospital admission (readmission studies), planned admissions, and transfers from other units or hospitals; mean sample age (>70 vs. ≤70); country (US/other); selection of hospital units versus inclusion of entire hospitals; publication year (>1992 vs. ≤1992); and inclusion versus exclusion of indirect DRM. Information for the last study characteristic was retrieved from the reported types of drug-therapy problems in each study. For example, studies that reported underdoses or lack of drug therapy as problems were considered to include indirect DRM. Overall study quality was not formally assessed, but we explored the association between prevalence estimates and 3 selected methodologic differences in PDRA ascertainment, assuming the following 3 characteristics would be more desirable: case assessments that included physician or patient interviews versus assessments that relied on medical chart review alone, assessments by ≥2 reviewers,

Table 2. Comparison of Selected Study Characteristics Characteristic

Category

Reference

Hospital units

entire

8-10, 13, 14, 16-18 (except psychiatric unit)

selected

7, 20, 21: ICU; 11: ICU and internal medicine; 12: acute medical; 15: acute geriatric and medical ward at heart care unit; 19: ICU and oncology

included

8-14, 16-18

excluded

7, 15, 19-21

implicit

16

explicit (criteria)

7 and 21 (Karch et al.22); 10 and 12 (Hallas et al.23); 11 (Hallas et al.23 and Naranjo et al.24); 14 (Karch et al.22 and Hallas et al.23); 19 (own criteria); 20 (Kramer et al.25)

structured implicit (criteria)

8 (explicit for ADR: Naranjo et al.24); 9, 13 (explicit for ADR: Karch et al.22); 15, 17, 18

implicit

7-9, 15-18, 20, 21

explicitb

11-13, 19

structured implicit

10, 14

hospital medical records only

9, 12, 15, 17, 19-21

hospital MD and/or patient interview

7, 8, 10, 11, 13, 16 (additional GP interview); 14 (additional GP interview and blood sample analysis on admission); 18 (additional review of primary care records)

yes

consensus: 7, 9, 14, 16, 20, 21

Indirect DRMa

DRA assessment (association between therapy and admission)

Assessment of preventability

Data sources included in assessment

Multiple reviewers required for final judgment

independent: 8 (lowest category of 4 judgments), 10 (agreement of 2), 13 (2, if disagreement consensus by 7), 17, 18 (2 of 3) no (1 reviewer)

11, 12, 15, 19

ADR = adverse drug reaction; DRA = drug-related admission; DRM = drug-related morbidity; GP = general practitioner; ICU = intensive care unit. a Categorization based on reported drug therapy problems: underdosage, untreated indication, failure to receive prescribed drug, and therapeutic failure were considered indirect DRM. b Explicit criteria for preventability assessment can be found in Appendix I.

www.theannals.com


■



■

1241


and assessments done with explicit or structured implicit criteria versus implicit judgment. In each meta-regression model, the dependent variable was the transformed PDRA estimate and the independent variable was a specific study characteristic. These regressions were fit with random-effects weights (reciprocal sum of the estimated within-study and among-study variances) using restricted maximum likelihood estimation of the between-study variance to account for heterogeneity. The meta-regression models produced point estimates and 95% CIs of prevalence ORs comparing studies that differ with respect to each characteristic. For example, if investigations conducted in one way reported an average PDRA prevalence of 14.0% and those conducted in another way reported an average prevalence of 4.2%, the prevalence OR would be (0.14/0.86)/(0.042/0.958) = 3.7. Because of the explanatory and observational nature of systematic reviews, formal inferences are not drawn from test statistics, p values, and CIs. Prevalence of PDRAs Fifteen studies7-21 published between 1980 and 1999 met the inclusion criteria for PDRA studies (Table 1). Major reasons for rejecting an article identified by the key word search were scope too narrow (e.g., focus on specific indications or disease states), proportion of preventable cases not reported, or case identification based on spontaneous report or other methods of unknown validity. Two other studies36,37 that otherwise met the inclusion requirements were excluded because they were conducted in emergency departments (EDs) and counted ED visits rather than hospitalizations. All studies were conducted in industrialized countries: 8 in Europe, 4 in the US, 2 in Australia, and 1 in Canada. The 15 studies report a median DRA prevalence of 7.1% (IQR 5.7–16.2%) and a median PDRA prevalence of 4.3% (IQR 3.1–9.5%). Overall, the median preventability rate was 59% (IRQ 50–73%). Included reports used various terms to denote an undesirable event or outcome related to drug therapy, such as drug-related problems, iatrogenic disease, DRAs, ADE, or named specific problems such as admissions due to noncompliance and adverse drug reactions. The studies showed differences, especially in the definition of drug relatedness, by including or excluding indirect DRM (lack of access or effectiveness of drug therapy). Hallas et al.23,33-35 have published a number of studies conducted in different hospital units. We chose to include a summary report that estimates PDRA prevalence for the entire hospital rather than by hospital unit because it was more generalizable.14 Hallas et al. adjusted the summary PDRA prevalence of 3.4% for unrepresentative patient distributions due to equal sample sizes across units.

test (p = 0.5) suggest little or no evidence of publication bias. However, the individual study results (PDRA prevalence estimates) were highly heterogeneous (Cochran’s Q = 176; df = 14; p < 0.001). Consequently, an overall summary estimate using meta-analytic methods was not appropriate. Stratified Analysis and Meta-Regression Studies varied with regard to the types of admissions sampled (Table 1). Four studies included admissions and transfers to intensive care units (ICUs) only,7,11,20,21 4 exclusively sampled unplanned admissions and excluded transfers,8,9,13,18 and 2 sampled readmissions (assumed to be unplanned and no transfers). The remaining 5 studies reportedly sampled all admissions or provided insufficient information about exclusions. For the stratified analysis, we assumed that these 5 investigations included transfers and planned admissions if not stated otherwise. The inclusion or exclusion of first admissions (i.e., readmission) was the study characteristic most strongly associated with PDRA prevalence (Table 3). The 2 studies16,17 confined to readmissions reported an average PDRA prevalence estimate of 14.0%, approximately 4 times higher (estimated prevalence OR = 3.7) than the average estimate from the trials that included first admissions as well. This association showed little evidence of being appreciably confounded by other characteristics (e.g., age). PDRA estimates from the 2 studies were reasonably homogeneous (p = 0.6), but estimates from the remaining 13 stud-

Publication Bias and Heterogeneity The roughly symmetrical appearance of the funnel plot (Figure 1) and the high p value from the Begg–Mazumdar29 1242

■


■

Figure 1. Funnel plot of 15 estimates of the prevalence of preventable drugrelated hospital admissions.



www.theannals.com


ies remained highly heterogeneous (p < 0.001). The inclusion or exclusion of transfers or planned admissions showed little evidence for an association with prevalence estimates (OR ~ 1). The mean age of admissions was the next most strongly associated characteristic with PDRA prevalence. Studies of comparably older patients (mean age >70) reported prevalence estimates about twice as high as studies of younger patients (Table 3). The characteristic of indirect DRM indicates that adverse patient outcomes due to therapeutic failure or lack of access to drug therapy (e.g., omission errors) were considered as being drug-related. Five studies7,15,19-21 counted only directly DRAs (e.g., overdoses, adverse drug reactions), 3 of which were conducted in the 1980s (Table 2). Consistent with expectations, trials with a more comprehensive DRM definition that included indirect drug-related causes tended to report higher prevalence estimates (estimated PDRA prevalence OR = 1.9; Table 3).

The remaining characteristics listed in Table 3 had little or no apparent association with PDRA prevalence. To evaluate whether investigations performed in specific hospital units would have oversampled high-risk patients, we compared studies that sampled from specific hospital units with those that sampled from an entire hospital population. The prevalence OR of 1.1 does not support the hypothesis that studies conducted in specific units tended to include high prevalence samples, but the 95% CI for this estimate is wide (0.57–2.3) owing to the small number of studies. All studies used expert medical chart review to identify and assess PDRAs, although the review and judgment procedures differed (Table 2). Considered either individually ORs or together (≥2 of the more desirable ascertainment characteristics), the estimated ORs for the 3 characteristics of PDRA ascertainment were only 1.0 –1.5 (Table 3). Strong trends toward higher PDRA prevalence estimates were not associated with lower-quality studies, although the small number of studies led again to wide CIs. These data do not support the hypothesis that studies with weaker review procedures substantially overreported PDRA prevalence. Likewise, there was Table 3. Prevalence Estimates and Odds Ratios for little evidence of an association between Stratum-Specific PDRA Study Groups PDRA prevalence and country or year of publication (Table 3). Characteristic Category No. Average PDRA Prevalence Studies PDRAa (%) OR (95% CI)b The ability to address confounding of the crude prevalence ORs in the meta-regression First hospital excluded 2 14.0 3.7 (1.5 to 8.9) analysis was limited by the fact that only 2 admission included 13 4.2 studies were confined to readmissions, 2 others Mean age >70 6 7.6 2.0 (0.95 to 4.2) were missing information for age, and there ≤70 7 3.9 missing 2 was a limited number of studies. Therefore, CIs Indirect DRM included 10 6.1 1.9 (0.92 to 3.9) became wider when >1 characteristic was exexcluded 5 3.3 amined simultaneously, but the general direcCountry US 4 6.9 1.6 (0.73 to 3.6) tion of the estimated ORs remained the same. other 11 4.4 In summary, inclusion/exclusion of first admisTransfers from excluded 8 5.9 1.5 (0.71 to 3.0) sion, average sample age, and inclusion/excluother units or included 7 4.1 sion of indirect DRM seem to affect prevalence hospitals estimates, but are limited in explaining overall Planned excluded 6 6.0 1.3 (0.61 to 2.8) variation among the specific results. admissions included 9 5.0 Hospital units

entire hospital selected units

8 7

5.8 4.5

1.1 (0.57 to 2.3)

Year of publication

≤1992 >1992

8 7

5.0 5.0

1.0 (0.47 to 2.1)

MD/patient interviewc

yes no

8 7

5.8 4.2

1.4 (0.67 to 2.9)

Specified criteria yes for preventability no judgmentc

6 9

5.0 4.9

1.0 (0.47 to 2.2)

4 11

4.8 5.5

1.1 (0.49 to 2.7)

8 7

6.0 4.0

1.5 (0.75 to 3.1)

Multiple reviewersc

no yes

PDRA 2 or 3 yes ascertainment 0 or 1 yes (aggregate of 3 characteristicsc above)

Significance

DRM = drug-related morbidity; PDRA = preventable drug-related admission. a Estimates from random-effects meta-regression with no covariates. b Quoted to 2 significant digits. c Specified characteristics of PDRA ascertainment.

www.theannals.com

The results of this systematic review suggest that preventable DRAs represent a significant public health concern in ambulatory care. More than half of DRAs were not considered acceptable consequences of therapeutic risk–benefit considerations, but rather were thought to be caused by inappropriate care and medication errors. The studies reviewed here came from 7 nations and were conducted over a period of almost 20 years. The statistical evidence of high heterogeneity and the many potential sources of that heterogeneity would frustrate any attempt to estimate prevalence for a country or time period. None of the countries or study periods, however, seem to have been immune to PDRAs. The data in Table 2 give the impres-


■



■

1243


sion of a widespread and long-standing problem in the quality of drug-therapy management. According to the National Hospital Discharge Survey,38 there were 31.8 million admissions to US hospitals in 1998. The top 6 primary diagnostic categories (heart disease, delivery, neoplasms, pneumonia, psychosis, cerebrovascular disease) accounted for 3–12% of these admissions. Any of the studies in this article, or all combined, suggest that inappropriate management of drug therapy may be a leading cause of hospital admissions in developed countries. The IOM report1 suggested, “There is evidence indicating that adverse drug events account for a sizeable number of admissions to inpatient facilities.” This systematic review confirms the IOM report’s suspicion and suggests that preventable DRM in ambulatory care is at least as significant and prevalent as within inpatient care environments. Our findings show higher prevalence rates than 2 previous meta-analyses on DRAs.39,40 Einarson39 reviewed 36 studies published between 1966 and 1989. The estimated DRA prevalence was between 4.9% (median) and 5.1% (weighted meta-analytic estimate). His review did not require the assessment of preventability; therefore, many of the studies were excluded from our analysis. Our higher DRA estimate (7%) may, in part, be due to the expansion in the 1990s of the definition of drug-relatedness to include indirect DRM (e.g., lack of therapy), which constituted more than one-half of the identified cases in some recent studies. In 1998, Lazarou et al.40 estimated the proportion of admissions caused by adverse drug reactions to be 4.7%, which is smaller than our median DRA prevalence and larger than our median PDRA prevalence. The 2 reviews are largely complementary: only 3 studies were included in both. Lazarou focused on adverse reactions and, thus, intended to expose the inherent adverse effects of drug products, while we addressed the adverse consequences associated with inappropriate use of medicines. An objective of the Lazarou et al. study was “to show that there are a large number of serious adverse drug reactions even when drugs are properly prescribed and administered.” Their metaanalysis attempted to exclude adverse outcomes caused by overdoses, underdoses, or other errors, while we focused on injury that occurred because drug therapy was inappropriate. Our median preventability rate of 59% suggests that, according to chart reviews, the majority of DRAs were not seen as the result of an inherent and unavoidable risk of the drug product, but rather as the result of inappropriate care and medication errors. In conclusion, our study places patient injury caused by drugs in a larger context and suggests that the issues of preventability and medication errors may be an even more salient problem in health care than unpreventable idiosyncratic drug reactions. ASSOCIATION WITH SELECTED STUDY CHARACTERISTICS

The results of meta-regression allow interpretation of overall study findings within the context of these charac1244

■


■

teristics. Three major categories of study characteristics were considered: sample characteristics, applied definitions of drug-relatedness and preventability, and applied methods of case ascertainment. Differences in Sample Characteristics

Our review was designed to exclude studies that restricted their prevalence estimates to specific patient populations or disease states; nevertheless, we did include investigations that sampled only patients >65 years old because this age group represents a substantial proportion of hospitalized patients in developed countries (e.g., 39% of all discharges from short-stay hospitals in the US in 1998).38 Meta-regression suggested an association between older age and PDRA prevalence, which is in agreement with the findings of an inpatient study41 that assessed adverse consequences of inappropriate drug therapy. It should be noted, however, that age was an ecologic variable (i.e., only the average age of the total study sample was known). Ideally, age-specific PDRA prevalence estimates would be available from all reports. We also could not test for confounding, such as an association between older age and an increased number of diseases or prescribed drugs. Another sample characteristic that showed a strong association with PDRA prevalence was the exclusion of first hospital admissions (readmission studies). The higher prevalence estimates in the 2 studies16,17 excluding first hospital admissions suggest possible deficiencies in the management of patient discharge and the transition to ambulatory care (seamless care). This issue also should be considered when choosing sampling criteria for research. Other characteristics were tested for an association with prevalence estimates. In their inpatient studies, Bates et al.42 and Cullen et al.43 reported a twofold higher PDRM prevalence in ICUs than in general care units.11 Five of 7 PDRA studies in our analysis that selected specific units included ICUs, but meta-regression did not detect an association. Based on our findings, we conclude that the reviewed studies that sampled high-risk units, or excluded transfers or planned admissions did not overestimate PDRA prevalence. Finally, the majority of studies may have concerned admissions to teaching hospitals. Some44 have suggested that teaching hospitals may be associated with a higher prevalence of ADEs. We were not able to explore this association because we could not reliably distinguish all of the teaching and nonteaching hospitals in our sample. However, this association was suggested by an inpatient study44 describing mainly hospital-acquired DRM. In contrast, DRAs reflect ambulatory care, which may or may not lead to an admission to a teaching hospital. Applied Definitions and Terminology

Variation in what was considered to be drug-related emphasizes the need for clear definitions, not only for comparison across studies, but also to improve the reliability of reviewers’ judgments. For example, Hallas et al.23 empha-



www.theannals.com


sized the need for explicit criteria by describing the different views of a general practitioner and the investigators concerning the term drug-related. A case of insulin-induced hypoglycemia was allegedly not drug-related, but rather was “due to insufficient carbohydrate intake.” Most importantly, recent studies, including the IOM report,1 have expanded their definition of preventable DRM to include indirect drug-related causes, such as errors of omission or underdoses. Including indirect causes undoubtedly leads to higher prevalence rates, but it permits a more balanced and comprehensive assessment of drug therapy quality and its consequences. The reviewed studies showed substantial differences in the applied definitions of preventability (if provided at all), even though they used similar labels. For example, Darchy et al.7 defined nonavoidable as occurring even though “management [was] the best that medical science can provide.” Hallas et al.14 distinguished between “possibly avoidable by an effort exceeding the obligatory demands” and definitely avoidable as “resulting from a drug treatment procedure inconsistent with present-day knowledge of good medical practice.” In this context, judgment could also be influenced by the original objective of the studies. Wilson et al.,45 looking at ADEs during hospital stays, proposed that higher prevalence estimates could be explained by their focus on prevention, whereas the Harvard Medical Practice Study,46 with ostensibly identical study methods, was concerned with medical negligence and litigation. Applied Methods of PDRA Ascertainment

We addressed quality of the included studies in 2 ways. First, assuming expert case review is at present the most valid method for PDRA identification, we excluded studies that used unvalidated screening tools or voluntary reports to identify cases. However, only a few studies47-49 have compared different retrieval methods. One working group49 comparing chart review with screening instruments and voluntary reports from hospital staff concludes that no method is foolproof and comprehensive. In another study, Leape et al.50 compared physician chart reviews against records of the hospitals’ risk management and found that, in 8.9% of the cases, reviewers failed to identify negligence. These findings suggest that the overall prevalence may be higher than our findings suggest. Second, we identified 3 study characteristics that might affect the validity of case identification and final assessment. The characteristics failed to show a strong association with prevalence estimates, possibly because there was substantial variation within stratum-specific groups. We found no evidence that the use of explicit judgment criteria affected prevalence estimates.51 The number of required reviewers also did not show a strong association with prevalence estimates, which may be explained by the variation among consensus methods. For example, some studies required majority rules, while others described a “team decision.” The study8 with the smallest preventability ratio used the most conservative procedure by taking the lowest www.theannals.com

category of 4 judgments. Some investigators9,12,42 have noted that the quality of studies is impacted by the completeness of information available for review. We, therefore, compared studies that included prospective patient information (patient or MD interview) with those that relied on chart review alone. Stratum-specific estimates suggest lack of information may underestimate rather than overestimate PDRA prevalence. In summary, even though our analysis did not show an association, methods of PDRA ascertainment might affect study reliability and prevalence estimates. Opportunities for improvements exist in the implementation of standardized terminology and assessment methods, broader access to patient data and medication records, and the development of validated and more efficient tools for case identification. Limitations This systematic review may be incomplete because not every available database was searched, nor were authors contacted for unpublished data. Nevertheless, there was no evidence for publication bias that would indicate a study selection bias and biased findings. Second, the studies do not represent the universe of hospital admissions, and we do not know how widely our findings may be applied. However, we tested various characteristics that may have affected external validity and addressed them in the discussion of our findings. No trend toward overestimation of PDRA prevalence was observed. However, sample-size limitations did not allow us to compare more than 1 characteristic at a time, even though studies may combine several characteristics. Because characteristics were not randomly assigned, which is a limitation of any observational study, reported trends are only exploratory and not confirmatory in nature. Furthermore, stratum-specific comparisons were limited to the retrievable information from the articles, and other untested characteristics might help to more fully explain heterogeneity. Finally, this review only investigated consequences of inappropriate drug therapy in ambulatory care leading to hospital admission. Other healthcare settings, such as longterm care, or other clinical or surrogate adverse outcomes determined through data, such as physician office visits, could not be included because of the lack of available studies. Summary Evidence suggests that PDRAs reflect a significant problem, that is, inappropriate management of drug therapy in ambulatory care. Although we cannot precisely estimate PDRA prevalence for any specific population, we believe that the range found in this review, 3–9% of admissions depending on sample characteristics and applied definitions, is the best available evidence concerning the magnitude of this problem. More than 50% of DRAs were judged as preventable; therefore, medication errors and in-


■



■

1245


appropriate drug therapy appear at least as important as idiosyncratic drug reactions. The heterogeneity of reported prevalence estimates could not fully be explained by the reviewed study characteristics. Variation in study findings is partially explainable by the differences in applied definitions of drug-relatedness, sampling of older patients, and readmissions. No other evidence of overestimation of PDRA prevalences was observed. Pending further research, clinicians should be especially aware of the potential risk carried by elderly patients and by those who have recently been hospitalized. The inclusion of indirect drug-related causes for patient morbidity (errors of omission) was associated with higher prevalence estimates, but produces a more complete assessment of drug therapy quality and its impact on patient health. Similar to this study, the Harvard Medical Practice Study42 of in-hospital care found wide variation of ADE rates. Only part of the variation could be explained. The authors concluded that adverse effects and negligence are not randomly distributed, and that certain types of hospitals have significantly higher rates of injuries due to substandard care. The same may possibly be said of ambulatory care. Almut G Winterstein PhD, Clinical Assistant Professor, Department of Pharmacy Health Care Administration, College of Pharmacy, University of Florida, Gainesville, FL Brian C Sauer BSc, Graduate Student, Department of Pharmacy Health Care Administration, College of Pharmacy, University of Florida Charles D Hepler PhD, Distinguished Professor, Department of Pharmacy Health Care Administration, College of Pharmacy, University of Florida Charles Poole MPH ScD, Associate Professor, Department of Epidemiology, School of Public Health, University of North Carolina at Chapel Hill, Chapel Hill, NC Reprints: Almut G Winterstein PhD, Department of Pharmacy Health Care Administration, College of Pharmacy, University of Florida, PO Box 100496, Gainesville, FL 32610-0496, FAX 352/3927782, E-mail [email protected]

References 1. Committee on Quality of Health Care in America, Institute of Medicine: to err is human—building a safer health system. Washington, DC: National Academy Press, 1999. 2. Brennan TA. The Institute of Medicine report on medical errors—could it do harm? Tex Med 2000;96:13-5. 3. McDonald CJ, Weiner M, Hui SL. Deaths due to medical errors are exaggerated in Institute of Medicine Report. JAMA 2000;284:93-5. 4. Leape LL. Institute of Medicine medical error figures are not exaggerated. JAMA 2000;284:95-7 5. Sox HC Jr, Woloshin S. How many deaths are due to medical error? Getting the number right. Eff Clin Pract 2000;3:277-89. 6. Hayward RA, Hofer TP. Estimating hospital deaths due to medical errors: preventability is in the eye of the reviewer. JAMA 2001;286:41520. 7. Darchy B, Le Miere E, Figueredo B, Bavoux E, Domart Y. Iatrogenic diseases as a reason for admission to the intensive care unit: incidence, causes, and consequences. Arch Intern Med 1999;159:71-8. 8. Ng D, Gosh DG, Harris J, Whitehead C. Unplanned medication-related admissions to an acute care general teaching hospital. Aust J Hosp Pharm 1999;29:84-7. 9. Raschetti R, Menniti IF, Morgutti M, Belisari A, Rossignoli A. Adverse drug events in hospitalized patients. JAMA 1997;277:1351-2.

1246

■


■

Appendix 1. DRA Preventability Criteria Nelson and Talbert11 (abbreviated): A DRA was definitely avoidable if the patient (1) did not take a drug that is known to reduce or prevent the symptoms according to the prescribed directions, (2) had a known allergy, (3) had a disease for which the drug was contraindicated, and (4) took a drug that was not indicated. Possibly avoidable: failure to monitor by a physician at reasonable time intervals and inadequate monitoring due to inability to see a physician (e.g., financial difficulties). Courtman and Stallings12 (abbreviated): A DRA was avoidable if (1) drug treatment was obviously inappropriate or contraindicated, (2) no measures were taken to counteract known effects of the drug, or (3) patients were noncompliant or insufficiently educated about their medications. Possibly avoidable: if the patient’s disease state was considered to be potentially changing, thereby resulting in the need for altered drug therapy. Dartnell et al.13 (abbreviated): A DRA was avoidable whether or not (1) drug was indicated or contraindicated, (2) dosage differed from accepted recommendations, (3) adequate monitoring was provided, (4) adequate counseling was provided, (5) alternative or no drug therapy could have been undertaken, and (6) admission was likely regardless of (inappropriate) use of drug therapy. Lakshmanan et al.19: (1) a toxic effect where levels could have been checked or were available but ignored, (2) use of contraindicated drugs, and (3) failure to detect adverse effects that were present before admission. Immunologic or idiosyncratic reactions and predictable adverse effects that are not avoidable (e.g., chemotherapy-induced neutropenia) were not considered preventable. The authors distinguish between effects that evolved so rapidly that they could not have been prevented and reactions evolving over weeks to months “such that the victims could have been examined as outpatients and had their therapy altered.”

10. Cunningham G, Dodd TRP, Grant DJ, McMurdo MET, Richards RME. Drug-related problems in elderly patients admitted to Tayside hospitals, methods for prevention and subsequent reassessment. Age Ageing 1997;26:375-82. 11. Nelson KM, Talbert RL. Drug-related hospital admissions. Pharmacotherapy 1996;16:701-7. 12. Courtman BJ, Stallings SB. Characterization of drug-related problems in elderly patients on admissions to a medical ward. Can J Hosp Pharm 1995;48:161-6. 13. Dartnell JGA, Anderson RP, Chohan V, Galbraith KJ, Lyon ME, Nestor PJ, et al. Hospitalisation for adverse events related to drug therapy: incidence, avoidability, and costs. Med J Aust 1996;164:659-62. 14. Hallas J, Gram LF, Grodum E, Damsbo N, Brosen K, Haghfelt T, et al. Drug-related admissions to medical wards: a population-based survey. Br J Clin Pharmacol 1992;33:61-8. 15. Lindley CM, Tully MP, Paramsothy V, Tallis RC. Inappropriate medication is a major cause of adverse drug reactions in elderly patients. Age Ageing 1992;21:294-300. 16. Nikolaus T, Specht Leible N, Kruse WHH, Oster P, Schlierf G. [The early rehospitalization of elderly patients. Causes and prevention]. German. Dtsch Med Wochenschr 1992;117:403-7. 17. Bero LA, Lipton HL, Bird JA. Characterization of geriatric drug-related hospital readmissions. Med Care 1991;29:989-1003. 18. Bigby J, Dunn J, Goldman L, Adams JB, Jen P, Landefeld CS, et al. Assessing the preventability of emergency hospital admissions. A method for evaluating the quality of medical care in a primary care facility. Am J Med 1987;83:1031-6. 19. Lakshmanan MC, Hershey CO, Breslau D. Hospital admissions caused by iatrogenic disease. Arch Intern Med 1986;146:1931-4. 20. Trunet P, Borda IT, Rouget A-V, Rapin M, Lhoste F. The role of drug-induced illness in admissions to an intensive care unit. Intensive Care Med 1986;12:43-6.



www.theannals.com

Preventable Drug-Related Hospital Admissions 21. Trunet P, Le Gall JR, Lhoste F, Regnier B, Saillard Y, Carlet J, et al. The role of iatrogenic disease in admissions to intensive care. JAMA 1980;244:2617-20. 22. Karch FE, Lasagna L. Adverse drug reactions. JAMA 1975;234:123641. 23. Hallas J, Harvald B, Gram LF, Grodum E, Brosen K, Haghfelt T, et al. Drug related hospital admissions: the role of definitions and intensity of data collection, and the possibility of prevention. J Intern Med 1990;228:83-90. 24. Naranjo CA, Busto U, Sellers EM, Sandor P, Ruiz I, Roberts EA, et al. A method for estimating the probability of adverse drug reactions. Clin Pharmacol Ther 1981;30:239-45. 25. Kramer MS, Leventhal JM, Hutchinson TA, Feinstein AR. An algorithm for the operational assessment of adverse drug reactions. I. Background, description and instructions for use. JAMA 1979;242:623-32. 26. Ashton CM, Kuykendall DH, Johnson ML, Wray NP. An empirical assessment of the validity of explicit and implicit process-of-care criteria for quality assessment. Med Care 1999;37:798-808. 27. Shadish WR, Haddock CK. Combining estimates of effect size. In: Cooper H, Hedges LV, eds. The handbook of research synthesis. New York: Russel Sage Foundation, 1994:266. 28. Greenland S, Rothman KJ. Introduction to categorical statistics. In: Rothman KJ, Greenland S, eds. Modern epidemiology. Philadelphia: Lippincott-Raven, 1998:240-1. 29. Begg CB, Mazumdar M. Operating characteristics of a rank correlation test for publication bias. Biometrics 1994;50:1088-101. 30. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;7:177-88. 31. Sharp SJ. Meta-analysis regression. Stata Tech Bull 1998;42:16-22. 32. Thompson SG, Sharp SJ. Explaining heterogeneity in meta-analysis: a comparison of methods. Stat Med 1999;18:2693-708. 33. Hallas J, Haghfelt T, Gram LF, Grodum E, Damsbo N. Drug-related admissions to a cardiology department; frequency and avoidability. J Intern Med 1990;228:379-84. 34. Hallas J, Jensen KB, Grodum E, Damsbo N, Gram LF. Drug-related admissions to a department of medical gastroenterology. The role of selfmedicated and prescribed drugs. Scand J Gastroenterol 1991;26:174-80. 35. Hallas J, Worm J, Beck-Nielsen J, Gram LF, Grodum E, Damsbo N. Drug-related events and drug utilization in patients admitted to a geriatric hospital department. Dan Med Bull 1991;38:417-20. 36. Tafreshi MJ, Melby MJ, Kaback KR, Nord TC. Medication-related visits to the emergency department: a prospective study. Ann Pharmacother 1999;33:1252-7. 37. Dennehy CE, Kishi DT, Louie C. Drug-related illness in emergency department patients. Am J Health Syst Pharm 1996;53:1422-6. 38. Hall MJ, Popovic JR. 1998 summary: national hospital discharge survey. Advance Data, NCHS 2000;316:1-17. 39. Einarson TR. Drug-related hospital admissions. Ann Pharmacother 1993;27:832-40. 40. Lazarou J, Pomeranz BH, Corey PN. Incidence of adverse drug reactions in hospitalized patients. JAMA 1998;279:1200-5. 41. Leape LL, Lawthers AG, Brennan TA, Johnson WG. Preventing medical injury. Qual Rev Bull 1993;19:144-9. 42. Bates DW, Cullen DJ, Laird N, Petersen LA, Small SD, Servi D, et al. Incidence of adverse drug events and potential adverse drug events: implications for prevention. JAMA 1995;274:29-34. 43. Cullen DJ, Sweitzer BJ, Bates DW, Burdick E, Edmondson A, Leape LL. Preventable adverse drug events in hospitalized patients: a comparative study of intensive care and general care units. Crit Care Med 1997;25:1289-97. 44. Brennan TA, Hebert LE, Laird NM, Lawthers A, Thorpe KE, Leape LL, et al. Hospital characteristics associated with adverse events and substandard care. JAMA 1991;265:3265-9. 45. Wilson RM, Runciman WB, Gibberd RW, Harrison BT, Newby L, Hamilton JD. The Quality in Australian Health Care Study. Med J Aust 1995;163:458-71. 46. O’Neil AC, Petersen LA, Cook EF, Bates DW, Lee TH, Brennan TA. Physician reporting compared with medical-record review to identify adverse medical events. Ann Intern Med 1993;119:370-6. 47. Cullen DJ, Bates DW, Small SD, Cooper JB, Nemeskal AR, Leape LL. The incident reporting system does not detect adverse drug events: a problem for quality improvement. Jt Comm J Qual Improv 1995;21:5418.

www.theannals.com

48. Jha AK, Kuperman GJ, Teich JM, Leape LL, Shea B, Rittenberg E, et al. Identifying adverse drug events. JAMA 1998;5:305-14. 49. Hiatt HH, Barnes BA, Brennan TA, Laird NM, Lawthers AG, Leape LL, et al. A study of medical injury and medical malpractice. N Engl J Med 1989;321:480-4. 50. Karch FE, Smith CL, Kerzner B, Mazzullo JM, Lasagna L. Commentary—adverse drug reactions—a matter of opinion. Clin Pharmacol Ther 1976;19:489-92. 51. Leape LL, Brennan TA, Laird N, Lawthers AG, Localio AR, Barnes BA, et al. The nature of adverse events in hospitalized patients. Results of the Harvard Medical Practice Study II. N Engl J Med 1991;324:377-84.

EXTRACTO OBJETIVO: Estimar la prevalencia de admisiones hospitalarias prevenibles debido a errores en la medicación, y explorar si ciertas características de estudio afectan la información disponible sobre estimados de prevalencia. MÉTODO: Investigación de Medline (1966-December 1999), International Pharmaceutical Abstracts (1970-1999), e investigación manual usando los términos claves relacionadas al tópico en cuestión. Dos evaluadores revisaron independientemente los estudios seleccionados (publicados en revistas médica de reputación) para extraer información sobre estimados de prevalencia real y características de estudio. Para que los estudios fueran seleccionados, ellos tenían que evaluar las admisiones hospitalarias debidas a errores en la medicación. También, debían incluir una evaluación quantitativa de cómo se hubieran podido prevenir dichas admisiones hospitalarias. Se usó el análisis estratificado y meta-regresión para explorar la asociación entre las características de estudio y los estimados de prevalencia. SÍNTESIS: Cincuenta estudios reportaron una prevalencia media de admisiones hospitalarias prevenibles por errores en la medicación de 4.3% (rango interquartil de 3.1-9.5%). La tasa de preventibilidad media de admisiones por errores en la medicación fue de 59% (rango interquartil de 50-73%). No hubo evidencias de selección preferida de publicaciones de acuerdo al tamaño del estudio. Debido a que los resultados de los estudios individuales eran bastante heterogéneos (análisis estadístico Cochran’s Q = 176, grados de libertad = 14, p < 0.001), no se calculó el estimado del resumen meta-analítico. La inclusion o exclusión de las admisiones primarias fue la característica de estudio más directamente asociada a la prevalencia de admisiones hospitalarias prevenibles por errores en la medicación (estimado de prevalencia de 14%), siendo aproximadamente 4 veces mayor que el estimado promedio de los estudios que incluían también las admisiones primarias. La inclusión o exclusión de transferencias o admisiones planeadas mostró poca evidencia de asociación con los estimados de prevalencia. La edad promedio en el momento de la admisión hospitalaria fue la siguiente característica de estudio más directamente relacionada a la prevalencia de admisiones hospitalarias prevenibles por errores en la medicación. Otro factor de estudio que mostró estimados de prevalencia altos fue la morbididad debida a errores en la medicación indirecta, por ejemplo, errores de omisión o falla terapéutica. Hubo poca evidencia de otras asociaciones con estimados de prevalencia, por ejemplo, selección de unidas hospitalarias específicas, exclusión/inclusión de admisiones planeadas, el país, y características metodológicas para averiguar casos de admisiones hospitalarias prevenibles debido a errores en la medicación. CONCLUSIONES: La morbididad relacionada a errores en la medicación es un problema sanitario significantivo y en muchos casos prevenible. Los métodos de estudio sobre prevalencia varían y deberían considerarse cuando se interpretan resultados o se planean investigaciones futuras.

Encarnación C Suárez RÉSUMÉ

Estimer la prévalence des admissions hospitalières dues à des médicaments et rechercher si les caractéristiques sélectionnées dans les études affectent l’estimation de cette prévalence.

OBJECTIF:


■



■

1247

AG Winterstein et al. MÉTHODES: Recherche par mots-clés sur Medline (1966-1999), International Pharmaceutical Abstracts (1970-1999), et recherche manuelle. Deux lecteurs critiques ont sélectionné des études publiées dans des revues scientifiques et ont extrait les estimations de prévalence et les caractéristiques de l’étude. Les études devaient s’adresser aux conséquences de la prise médicamenteuse responsables de l’admission à l’hôpital, et inclure une évaluation quantifiée de leur évitabilité. Les analyses stratifiées et la méta-régression ont été utilisées pour explorer l’association entre les caractéristiques de l’étude et les estimations de prévalence. SYNTHÈSE DES DONNÉES: Quinze études ont rapporté une prévalence médiane d’admissions évitables de 4.3% (IC 3.1–9.5%). L’évitabilité moyenne des admissions hospitalières dues au médicament était de 59% (IC de 50–73%). Il n’a pas été possible de déterminer un biais de publication dû à la taille de l’étude. L’importante hétérogénéité des résultats individuels des études rendait inutile toute ré évaluation par méta analyse. L’analyse stratifiée a suggéré une association entre les estimations de prévalence et 3 caractéristiques des études: exclusion de la première admission (avec un taux moyen de prévalence de 14% en

réadmission, 4 fois plus élevé que si l’on inclut les premières admissions); âge moyen des admissions de plus de 70 ans comportant une fréquence 2 fois plus élevée que les séries incluant des sujets moins âgés; et inclusion d’une morbidité due indirectement au médicament, comme les erreurs ou les échecs thérapeutiques. D’autres associations entre caractéristiques et prévalence étaient moins bien démontrées, comme la sélection de services hospitaliers spécifiques, l’exclusion ou non des admissions programmées, le pays, et les caractéristiques méthodologiques de déclaration des cas d’admission hospitalière évitable dues au médicament. CONCLUSIONS: la morbidité causée par les médicaments est un problème significatif de santé publique dans les soins ambulatoires, et une proportion importante peut en être évitée. La diversité des méthodes d’étude de leur prévalence doit être prise en compte pour en interpréter les résultats comme pour planifier des recherches à venir. Jean-Marie Kaiser

Finding Strength in Weakness: A Study of Tribulation and Our Appropriate Response By William D Black MD ISBN 1-57921-3580-8 / Paperbound / xxii + 256 pp. / 2001 / $13.99 Dr. Black writes, “God allows [human suffering] in order to bring about a more ultimate good in our lives. Any response to tribulation that moves us farther away from God should be considered sin. “The key to successfully moving through difficult circumstances lies within the human spirit. There is nothing that quite satisfies the longings of our spirits like the God who created them. In this life, we may never completely understand why we are undergoing tribulation, but we have ample evidence of how we are to handle it. Jesus is our example, and we are supposed to walk as He did.” “Dr Black writes from the pathos of the examination room…From a lifetime of immersion in Holy Scripture, intermixed with his own multiplex of personal setbacks and devastating family illnesses, he brings us a book that the anguished and heartbroken will turn to again and again for solid help.” — Rev. Ronald L Siegenthaler, Coral Ridge Presbyterian Church, Ft. Lauderdale, FL “It is a fine book, bringing together a great deal of biblical wisdom that to my knowledge is not to be found between any other two book covers. It is a powerful piece of writing, and will surely do its readers a great deal of good.” — JI Packer, Regent College, Vancouver, BC, Canada For more information, visit www.FindingStrength.com or call toll-free 877-421-7323

1248

■


■



www.theannals.com