Sensitivity of Three Widely Used Questionnaires for Measuring ...

4 downloads 161 Views 547KB Size Report
This is the accepted manuscript version for posting on personal website. .... mean scores between the two groups by the pooled standard deviation (PSD) (24, ...
TITLE:  Sensitivity  of  Three  Widely  Used  Questionnaires  for  measuring  Psychological  Distress   among  patients  with  Type  2  Diabetes  Mellitus   Tan  LSM,  1  Khoo  EYH,  2  Tan  CS,1  Griva  K,  3  AMIR  Mohamed  3,  NEW  Michelle  2,  Lee  YS,  4  Lee   Jeannette,  1  Tai  ES,  1,2    Wee  Hwee-­‐lin  5,6   1.  

School  of  Public  Health,  Yong  Loo  Lin  School  of  Medicine,  Singapore  

2.  

Department  of  Medicine,  National  University  Health  System,  Singapore  

3.  

Department  of  Psychology,  National  University  of  Singapore,  Singapore  

4.  

Department  of  Pediatrics,  Yong  Loo  Lin  School  of  Medicine,  Singapore  

5.  

Department  of  Rheumatology  &  Immunology,  Singapore  General  Hospital,  Singapore  

6.  

Department  of  Pharmacy,  National  University  of  Singapore,  Singapore  

  For  correspondence:   Assistant  Professor  Wee  Hwee  Lin   Department  of  Pharmacy   National  University  of  Singapore   10  Kent  Ridge  Crescent   Singapore  119260   Singapore     Tel:  65-­‐  6516-­‐5530   Fax:  65-­‐  6778  5698   e-­‐mail:  [email protected]    

This is the accepted manuscript version for posting on personal website. The final publication is available at http://link.springer.com/article/10.1007/s11136-014-0747-z    

 

1  

 

ABSTRACT     Background:   Although   a   range   of   generic   and   diabetes   specific   instruments   are   available   to   assess   emotional   distress,   no   studies   have   evaluated   sensitivity   in   relation   to   sample   size   requirement.   The   present   study   sets   out   to   compare   the   sensitivity   among   the   Diabetes   Health   Profile   psychological   distress   scale   (DHP   –   PD),   Problem   Areas   in   Diabetes   (PAID)   and   Kessler-­‐10   Psychological   Distress   scale  (K10).  We  hypothesized  that  the  diabetes-­‐specific  measures  (DHP-­‐PD  and  PAID)  would  require   smaller  sample  sizes  than  the  generic  measure  (K10)  yet  remain  specific.     Research  Design:  A  total  of  208  patients  with  type  2  diabetes  mellitus  (mean  age  45.2  (12.4)  years;   63.1%  males,  45.8%  Chinese,  11.3%  Malay  and  26.6%  Indian),  recruited  from  a  Singapore  tertiary   hospital  diabetes  clinic,  completed  the  English  DHP-­‐PD,  PAID  and  K10.    Clinical  information  derived   from  medical  records  and  HbA1c  was  recorded.  Effects  sizes  (ES),  ratio  of  ES  and  sample  size   requirement  relative  to  the  most  sensitive  questionnaire  were  computed.   Results:  A  comparison  of  patients  with  good  versus  poor  glycaemic  control  (HbA1c≥7.0)  revealed   that  using  K10  will  require  4  times  the  sample  size  of  a  study  using  the  PAID  in  order  to  detect  the   same  level  of  psychological  distress.    The  DHP-­‐PD  and  PAID  had  similar  sensitivity  when  comparing   between  patients  with  good  versus  poor  glycaemic  control.   Conclusions:    As  hypothesised,  sample  size  requirement  is  largest  for  K10  and  remarkably  similar  for   PAID  and  DHP-­‐PD.  This  information  is  useful  for  designing  clinical  trials  and  studies.    

 

2  

 

INTRODUCTION   Diabetes   Mellitus   (DM)   is   a   chronic   disease   associated   with   psychological   distress   such   as   withdrawal,   anger,   anxiety   and   depression   (1-­‐3).     Prevalence   of   psychological   distress   in   patients   with  DM  range  from  18%  to  52%   (1,  4)  globally  and  is  approximately  30%  (5,  6)  in  Singapore.  Two   forms   of   psychological   distress   exist   for   patients   with   DM,   namely   depressive   symptoms   (DS)   and   DM-­‐related   distress   (DRD).     DS   is   associated   with   clinical   depression,   encompassing   sadness,   frustration,   anxiety,   and   other   negative   mood   states.     Generic   questionnaires   are   often   used   as   screening  tools  to  identify  DS.    Kessler-­‐10  Psychological  Distress  scale  (K10)  is  one  such  questionnaire   that  is  growing  in  popularity  globally  (7-­‐9).    DRD,  on  the  other  hand,  is  attributed  to  the  burden  of   living   with   DM.   It   encompasses   physician-­‐related   stress,   treatment-­‐related   stress,   interpersonal   distress,   emotional   burden   of   disease,   food-­‐related   problems,   etc.     Only   DM-­‐specific   questionnaires,   which   are   targeted   to   identify   distress   related   to   DM   can   be   used   to   identify   such   distress.     Two   examples  of  DM-­‐specific  questionnaires  are  the  Diabetes  Health  Profile  (DHP)  and  Problem  Areas  in   Diabetes  (PAID).       While   reliability   and   validity   are   typical   considerations   for   selecting   a   questionnaire,   a   less   commonly   used,   yet   important   criteria,   is   sample   size   requirement   (sensitivity).     Sensitivity,   here,   describes   the   ability   of   an   instrument   to   detect   differences   between   populations   or   subgroups.     Hence,   a   sensitive   questionnaire   can   efficiently   differentiate   between   subgroups,   leading   to   significant   reduction   in   the   number   of   subjects   but   yet   detect   meaningful   results,   which   is   especially   useful  for  clinical  trials.   The   aim   of   our   study   is   to   compare   the   sensitivity   of   three   questionnaires   (DHP,   PAID   and   K10)   in   discriminating   psychological   distress   among   subgroups   of   patients   with   type   2   diabetes   mellitus  (T2DM).  In  particular,  a  comparison  of  DHP  and  PAID  will  aid  clinicians  in  making  a  choice   between   these   two   widely   used   diabetes-­‐specific   questionnaires.   We   hypothesised   that   the   DM-­‐ specific  measures  (DHP-­‐PD  and  PAID)  would  require  smaller  sample  sizes  than  the  generic  measure   (K10),  yet  retain  their  specificity.       3  

 

  METHODS   Study  Design  and  Participants   This  study  was  approved  by  the  National  Healthcare  Group  Domain  Specific  Review  Board   (Protocol  No.:  2011/02018).    Multi-­‐ethnic  patients  with  T2DM  between  21  and  65  years  old,  with  at   least  one  year  of  being  diagnosed  with  DM  were  recruited  from  the  diabetes  clinic  at  the  National   University  Hospital,  Singapore,  from  2011  to  2012.  Patients  were  selected  by  convenience  sampling   at  the  clinic  waiting  area.  Only  English  literate  patients  were  included  in  the  study.  Excluded  patients   consisted  of  those  with  severe  heart,  kidney,  and  liver  disease,  those  who  were  mentally   incapacitated,  and  alcohol  or  drug  abusers.  Informed  consent  was  obtained  from  all  participating   patients.    In  the  analysis,  we  only  presented  data  on  T2DM  patients  (Figure  1).     Data  collection   Data  on  demographic  factors  were  collected  using  self-­‐administered  questionnaires.  Ethnic   group  was  classified  as  Chinese,  Malay,  Asian  Indian  or  Others.    Marital  status  was  classified  as   “never  married”,  “currently  married”  or  “separated/divorced/widowed”.  Education  level  was   determined  based  on  the  number  of  schooling  years  and  was  categorized  into  10   years.       Psychological  distress  questionnaires   Kesseler-­‐10  Psychological  Distress  scale  (K10)   The  K10  (10)  is  a  generic  questionnaire,  consisting  of  10  items  designed  to  measure  the  level   of  distress  and  severity  associated  with  psychological  symptoms  in  population  surveys.    Each  item  in   K10  is  scored  1  to  5  (“None  of  the  time”  to  “All  of  the  time”).    Item  responses  are  summed  to   produce  an  overall  score.    It  is  popular  worldwide  because  the  instrument  is  short,  simple  to   administer,  had  been  validated  (7)  and  is  used  in  the  world  mental  health  survey  (11).       4  

 

Diabetes  Health  Profile  (DHP)   The  DHP-­‐18  items  (DHP-­‐18),  was  adapted  from  the  DHP-­‐1  (12)  to  identify  psychosocial   dysfunction  among  non-­‐insulin-­‐dependent  patients  (13).  It  consists  of  18  items  covering  three   dimensions:  psychological  distress  (DHP-­‐PD),  barriers  to  activity  (DHP-­‐BTA)  and  disinhibited  eating   (DHP-­‐DE).    Each  item  in  DHP-­‐18  is  scored  0  to  3  (“Never”  to  “Very  much”).    Item  responses  within   each  subscale  are  summed  to  produce  a  score  for  the  respective  subscales  (14). A  manuscript   reporting  the  validity  and  reliability  of  DHP-­‐18  in  Singapore  has  been  submitted  (15).    Only  the  DHP-­‐ PD  subscale  (containing  6  items)  was  used  in  this  analysis.       Problem  Areas  in  Diabetes  (PAID)   The  PAID  (16)  is  a  commonly  used  instrument  for  mapping  diabetes-­‐related  problem  areas.   It  comprises  20  items  covering  frequently  reported  emotional  states.  Each  item  in  PAID  is  scored  0  to   4  ("Not  a  problem"  to  "Serious  Problem").  The  sum  of  the  items  is  multiplied  by  1.25  to  yield  a  final   score  of  0-­‐100  (17).  The  PAID  had  been  validated  globally  (18)  as  well  as  in  Singapore  (19).       In  all  three  aforementioned  questionnaires,  higher  scores  indicate  higher  levels  of   psychological  distress.         Definitions   Glycemic  control  was  determined  by  measured  glycated  hemoglobin  (HbA1c).    We  classified   patients  with  good  control  of  HbA1c  as  achieving  the  standard  of  ≤7.0%  (20).  Medications  were   classified  into  treatment  types,  such  as  oral,  insulin  or  both.    Presenteeism,  which  is  a  measure  of   effectiveness  of  an  individual  who  goes  to  work  despite  having  an  illness,  was  measured  using  a   single  question  “On  a  scale  of  0  to  10,  how  effective  are  you  at  work?”    We  arbitrarily  considered  a   score  ≤5  (mid-­‐point  of  the  scale)  as  not  effective  at  work.    The  Family  Functioning  Measure  (FFM)  is  a   1-­‐item  question  used  to  determine  the  level  of  family  support  the  patient  is  getting.    The  item   response  ranges  from  1  “Poor”  to  5  “Excellent”.    We  considered  a  score  ≥3  as  having  the  presence  of   family  support  (21).     5  

 

  Statistical  analysis   The  missing  item  scores  were  imputed  where  possible  based  on  the  recommendations  by   the  developers  of  each  instrument  (17,  22,  23).  Participants  with  missing  overall  DHP-­‐PD,  PAID  or   K10  scores  were  excluded  list  wise  from  the  analysis  (Figure  1).    Mean  and  standard  deviations  were   used  to  describe  continuous  variables  while  percentages  were  used  on  categorical  variables.     Sensitivity   Sensitivity   was   determined   by   effect   size   (ES),   derived   by   dividing   the   differences   in   the   mean  scores  between  the  two  groups  by  the  pooled  standard  deviation  (PSD)  (24,  25).    Based  on  our   literature   review   of   Pubmed   using   terms   “type   2   diabetes   mellitus”,   “Diabetes   Health   Profile”,   “Problem   Areas   In   Diabetes”   and   “Kessler-­‐10”,   the   minimum   important   difference   (MID)   had   not   been   determined   for   any   of   the   instruments   used   in   the   study.     However,   several   studies   had   shown   that  the  ES  across  various  patient  sub-­‐groups  had  ranged  between  0.3  and  0.5  (24,  26).    Hence  we   decided  to  use  Cohen’s  ES  of  0.3  to  be  the  minimum  effect  size  to  detect  a  MID  for  this  study  (25).       Relative  effect  size  (RES)  was  used  to  determine  the  sensitivity  of  the  DHP-­‐PD  compared  to   PAID  and  K10  as  well  as  PAID  to  K10.    RES  was  expressed  as  the  ES  of  DHP-­‐PD  over  ES  of  PAID  and   K10  separately  as  well  as  the  ES  of  PAID  over  ES  of  K10.    RES  values  greater  (less)  than  1  will  suggest   that  the  DHP-­‐PD  was  more  (less)  sensitive  than  PAID  or  K10,  or  the  PAID  was  more  (less)  sensitive   than  K10,  in  discriminating  the  level  of  psychological  distress  between  various  sub-­‐groups.    Relative   sample  size  was  determined  from  the  RES,  defined  as  the  square  of  the  RES  (27).  Sub-­‐groups  based   on  clinical,  demographic  and  socio-­‐economic  and  social  functioning  variables  were  evaluated  on  the   basis  that  they  would  allow  for  identification  of  patients  who  may  benefit  from  tailored  programmes   to  reduce  psychological  distress,  e.g.,  gender-­‐specific  programmes.     Simulation    

To  determine  the  power  of  the  different  instruments  under  various  scenarios  in  a  systematic  

fashion  and  to  determine  the  generalizability  of  findings  from  the  data  analyses,  we  conducted  1000  

6  

 

simulations  iterations  on  the  DHP-­‐PD,  K10  and  PAID  scores  based  on  the  empirical  distribution  of  the   individual  scores  from  the  data  collected.    We  varied  the  following  factors:   1)  

Effect  sizes  using,  0.2  as  a  small  effect,  0.5  as  moderate  effect  and  0.8  as  a  large  effect  

2)  

Sample  size  of  the  smaller  group  with  n=10,  30,  50  and  100  

3)  

Ratio  of  the  sample  size  of  the  larger  group  to  the  smaller  group  with  the  following  specified   values:  1,  1.5,  2,  2.5.  

Based  on  the  variations  of  these  three  factors,  we  obtained  48  possible  combinations  for  each  of  the   three  instruments.    We  counted  and  tabulated  the  percentage  and  number  of  times  the  individual   instruments  (untied)  had  the  highest  power  to  detect  a  difference  in  scores.    Higher  power  is   preferred  because  this  suggests  the  instrument  is  more  likely  to  produce  a  statistically  significant   result  under  the  alternative  hypothesis  given  the  same  sample  size.    Simulation  was  performed  in  R,   version  3.0.3  (http://www.r-­‐project.org/),  using  functions,  such  as,  sample  and  t.test  (28).    All  other   analyses  were  performed  in  Stata  version  12  (29).  

 

  RESULTS    

Table  2  describes  the  socio-­‐demographic,  co-­‐morbidity  and  DM  control  of  208  patients  with  

T2DM  included  in  the  study  (Figure  1).    The  mean  (SD)  age  of  the  population  was  45.5  (11.9)  years   with  63%  males.  50%  of  the  subjects  were  Chinese,  followed  by  Indian  (28%)  and  Malay  (12%).  Most   of  the  subjects  had  at  least  one  co-­‐morbidity  with  retinopathy  and  cardiovascular  disease  at  13%,   followed  by  nephropathy  (8%)  and  neuropathy  (7%),  to  name  a  few.    71%  of  the  subjects  had  poor   control  of  their  disease  (i.e.,  HbA1c  ≥7.0%).   Sensitivity   As  expected,  all  measures  of  psychological  distress  were  sensitive  in  detecting  significant   differences  between  patients  with  co-­‐morbidities  versus  those  without  (ES>0.3),  with  the  exception   of  K10  in  detecting  those  with  retinopathy,  cardiovascular  and  cerebrovascular  disease  comorbidity   and  both  DM-­‐specific  questionnaire  in  detecting  patients  with  anaemia  comorbidity  respectively   (Table  3).    The  K10  detected  significant  differences  in  psychological  distress  between  males  and  

7  

 

females  (ES=0.39).    While  only  PAID  was  sensitive  in  detecting  psychological  distress  across   educational  status  (ES=0.32),  all  instruments  were  sensitive  when  comparing  across  sub-­‐groups  of   socio-­‐economic  status  (housing  type  and  household  income)  of  patients.    Of  the  three  variables   available  to  measure  modifiable  social  determinants  (presenteeism,  effective  outside  work  and  FFM),   we  found  that  the  DHP-­‐PD  was  most  sensitive  (ES>0.3  for  all  3  variables)  in  detecting  significant   differences  in  psychological  distress,  followed  by  PAID  (ES>0.3  for  2  out  of  3  variables).    K10  was  not   sufficiently  sensitive  in  detecting  any  significant  differences  among  the  modifiable  social   determinants.   RES  results  between  each  questionnaire  pair  and  ratio  of  sample  size  requirements  relative   to  the  most  sensitive  instrument  are  depicted  in  Table  4.    Generally,  across  most  subgroups,  DHP-­‐PD   and  PAID  tended  to  be  more  sensitive  than  K10  at  detecting  psychological  distress.    For  example,   when  comparing  between  patients  with  good  versus  poor  glycaemic  control,  the  RES  between  DHP-­‐ PD  and  K10,  PAID  and  K10  and  DHP-­‐PD  and  PAID  were  1.22,  1.67  and  0.73,  respectively.    In  terms  of   sample  size  requirement,  if  a  study  uses  the  K10  questionnaire,  it  will  require  2.8  times  the  sample   size  of  a  study  using  the  PAID  to  detect  the  same  level  of  psychological  distress.  However,  there   were  two  exceptions  where  the  K10  was  more  sensitive  at  detecting  psychological  distress:  1)   among  patients  with  anemia  co-­‐morbidities;  and  2)  in  some  demographic  and  socio-­‐economic   variables  (specifically  age,  gender,  marital  status  housing  type  and  household  income).     Simulation  

 

Of  the  three  instruments,  DHP-­‐PD  had  the  highest  power  the  most  number  of  times  (15  out   of  41  combinations),  while  PAID  and  K10  tied  in  13  out  of  41  combinations  (Table  5).  There  were   combinations  where  the  instruments  were  tied  for  the  highest  power,  i.e.at  least  two  of  the   instruments  had  exactly  the  same  power,  indicating  that  the  instruments  were  no  different  from   each  other  and  hence  no  instrument  was  deemed  superior  than  the  other  two  instruments.   However,  these  were  rare  and  hence  not  reported  (all  three  instruments  were  tied  in  5  out  of  48   combinations  and  two  instruments  were  tied  in  2  combinations).   8  

 

 

When  we  varied  the  effect  size,  at  low  levels  of  effect  size  (ES=0.2),  DHP-­‐PD  most  frequently  

reported  the  highest  power  (43.8%  of  the  time).  At  larger  effect  sizes  (ES=0.5)  DHP-­‐PD  and  K10  tied   with  each  reporting  the  highest  power  37.5%  of  the  time.   At  small  sample  sizes  (n=10),  DHP-­‐PD  and  K10  were  seen  to  have  equal  numbers  of  highest   power  (41.7%  of  the  time).    However,  as  the  sample  size  increases  to  n=30,  we  found  that  the  PAID   and  DHP-­‐PD  instruments  were  more  likely  to  report  highest  power  (45.5%  and  36.4%  respectively).   The  simulation  results  suggest  that  when  the  sample  sizes  increase  further,  we  find  no  differences   between  the  DM-­‐specific  instruments  from  the  generic  instrument,  suggesting  that  both  types  of   instruments  would  be  able  to  detect  significant  levels  of  psychological  distress  when  the  sample  size   is  large.     DISCUSSION   As  hypothesized,  in  this  first-­‐of-­‐its-­‐kind  study  that  concurrently  evaluated  three   questionnaires  of  psychological  distress,  the  DHP-­‐PD  and  PAID  were  generally  more  sensitive  than   K10  at  detecting  psychological  distress  among  sub-­‐groups  of  patients  with  T2DM.    This  translates  to   having  to  recruit  fewer  patients.       There  were  two  instances  where  the  K10  was  observed  to  be  more  sensitive  than  the  DHP-­‐ PD  and  PAID,  the  first  being  comparisons  between  patients  with  and  without  anaemia.  This  may  be   attributed  to  the  symptoms  of  anaemia  being  closely  associated  with  the  symptoms  of  diabetes,   hence  the  DM-­‐specific  instruments  might  not  be  adequate  in  detecting  psychological  distress  due  to   anaemia  (30).  The  second  instance  was  observed  across  subgroups  defined  by  socio-­‐demographic  or   socio-­‐economic  variables.  Age,  gender  and  marital  status,  housing  type  and  household  income  are   not  disease-­‐specific  variables.  Hence  it  is  not  surprising  that  K10  was  better  at  detecting   psychological  distress  between  the  various  categories.     PAID  was  more  sensitive  at  detecting  psychological  distress  with  regards  to  the  clinical   variables  but  was  equally  as  sensitive  as  the  DHP-­‐PD  where  social  functioning  variables  were   9  

 

concerned.  This  may  be  partly  explained  by  the  larger  number  of  clinical  items  in  PAID  compared   with  DHP-­‐PD.  However,  it  should  be  noted  that  DHP-­‐PD  is  not  a  stand–alone  questionnaire.    Thus,  by   just  including  the  DHP-­‐PD  sub-­‐scale,  we  have  missed  the  benefits  of  measuring  two  other  aspects  of   DM-­‐related  well-­‐being.  Nonetheless,  given  that  this  paper  was  focusing  on  detecting  psychological   distress,  it  was  necessary  for  us  to  extract  the  relevant  sub-­‐scale  to  ensure  equivalent  comparisons.             A  sensitive  questionnaire,  which  can  detect  differences  between  groups,  can  sometimes  be   used  to  describe  the  responsiveness  of  a  questionnaire,  which    can  detect  significant  changes  across   time  (31).    This  is  particularly  useful  as  most  studies  evaluating  psychometric  properties  of  patient   reported  outcome  questionnaires  are  cross-­‐sectional  rather  than  longitudinal  in  nature.    To  date,  of   the  three  questionnaires  analysed  in  our  study,  only  the  PAID  had  been  specifically  tested  for   responsiveness  (32).    Based  on  the  findings  of  our  study,  it  is  reasonable  to  extrapolate  that  DHP  will   exhibit  responsiveness  while  K10  would  not.       Based  on  the  simulation  results  stratified  by  ES  in  Table  5,  DHP-­‐PD  is  frequently  having  the   highest  power  when  the  ES  is  small  (ES=0.2)  while  PAID  is  frequently  having  the  highest  power  when   the  ES  is  large  (ES=0.8).  The  findings  from  simulation  and  data  analysis  thus  support  the   recommendation  to  use  DM-­‐specific  psychological  distress  questionnaires  (DHP-­‐PD  and  PAID)  over   K10  because  of  better  power  and  sensitivity  properties.  It  should  be  mentioned  that  there  are   scenarios  where  K10  would  have  the  highest  power,  e.g.,  when  the  sample  sizes  between  the  two   instruments  are  equal  (i.e.,  ratio  =  1).   Our  study  is  not  without  limitations.  First,  we  only  captured  English-­‐speaking  patients,  thus   limiting  the  generalizability  of  our  findings.  However,  based  on  the  Singapore  Census  2010,  75%  of   the  Singapore  resident  population  aged  25  to  65  was  English-­‐literate  (33).  Second,  the  Malays  were   slightly  under-­‐represented  in  our  patient  population.    Nonetheless,  we  managed  to  recruit  a  large   proportion  of  Indians,  which  was  predicted  to  be  the  most  likely  to  have  diabetes  of  the  three  ethnic   groups  (34).    Lastly,  the  presenteeism  item  was  dichotomised  based  on  the  mid-­‐point  of  the  scale   (≤5).      We  acknowledge  that  the  median  and  mean  values  were  generally  higher  than  the  mid-­‐point   10  

 

(8  and  7.5  out  of  10  points  respectively).    However,  we  decided  to  use  the  mid-­‐point  (≤5)  because  we   do  not  suspect  that  social  desirability  effect  would  be  present  due  to  the  following  reasons:  1)  the   response  scale  had  a  series  of  10  categories.    If  the  response  category  was  just  “yes/no”,  there   would  be  higher  level  of  response  bias  (35);  2)  the  questionnaire  was  self-­‐completed  and  anonymous,   hence  social  desirability  effect  would  be  negligible  (36).       CONCLUSION    

Overall,  based  on  the  results  from  the  data  analysis,  DM-­‐specific  psychological  distress  

questionnaires  (DHP-­‐PD  and  PAID)  were  more  sensitive  than  K10,  inferring  smaller  sample  size   requirements.  Separately,  PAID  was  more  sensitive  than  DHP-­‐PD  at  detecting  psychological  distress   among  subgroups  of  patients  with  T2DM.  Simulation  results  suggest  that  DHP-­‐PD  and  PAID  have   better  power  for  small  and  large  ES  respectively.  The  results  from  the  data  analysis  and  simulation   supports  the  usage  of  DM-­‐specific  psychological  distress  questionnaires  (DHP-­‐PD  and  PAID)  over  K10   because  of  better  sensitivity  and  power  properties.    In  conclusion,  our  study  has  presented   information  that  would  be  useful  in  helping  clinicians  and  researchers  decide  on  the  instruments  to   be  included  when  designing  clinical  trials  and  studies.             ACKNOWLEDGEMENTS   This  work  was  supported  by  the  grant  from  Ministry  of  Education  Singapore  Academic  Research   Fund  Tier  1  (Grant  No.:  FY2011-­‐FRC3-­‐007).     REFERENCES   1.   Anderson  RJ,  Freedland  KE,  Clouse  RE,  Lustman  PJ.  The  prevalence  of  comorbid  depression   in  adults  with  diabetes:  a  meta-­‐analysis.  Diabetes  care.  2001;24(6):1069-­‐78.  Epub  2001/05/26.   2.   Katon  WJ.  The  comorbidity  of  diabetes  mellitus  and  depression.  The  American  journal  of   medicine.  2008;121(11  Suppl  2):S8-­‐15.  Epub  2008/10/29.   3.   Pawaskar  MD,  Anderson  RT,  Balkrishnan  R.  Self-­‐reported  predictors  of  depressive   symptomatology  in  an  elderly  population  with  type  2  diabetes  mellitus:  a  prospective  cohort  study.   Health  and  quality  of  life  outcomes.  2007;5:50.  Epub  2007/08/09.   11  

 

4.   Ali  S,  Stone  MA,  Peters  JL,  Davies  MJ,  Khunti  K.  The  prevalence  of  co-­‐morbid  depression  in   adults  with  Type  2  diabetes:  a  systematic  review  and  meta-­‐analysis.  Diabetic  medicine  :  a  journal  of   the  British  Diabetic  Association.  2006;23(11):1165-­‐73.  Epub  2006/10/24.   5.   Chong  SA,  Subramaniam  M,  Chan  YH,  Chua  HC,  Liow  PH,  Pek  E,  et  al.  Depressive  symptoms   and  diabetes  mellitus  in  an  Asian  multiracial  population.  Asian  journal  of  psychiatry.  2009;2(2):66-­‐70.   Epub  2009/06/01.   6.   Verma  SK,  Luo  N,  Subramaniam  M,  Sum  CF,  Stahl  D,  Liow  PH,  et  al.  Impact  of  depression  on   health  related  quality  of  life  in  patients  with  diabetes.  Annals  of  the  Academy  of  Medicine,  Singapore.   2010;39(12):913-­‐7.  Epub  2011/01/29.   7.   Andrews  G,  Slade  T.  Interpreting  scores  on  the  Kessler  Psychological  Distress  Scale  (K10).   Australian  and  New  Zealand  journal  of  public  health.  2001;25(6):494-­‐7.  Epub  2002/02/05.   8.   Cairney  J,  Veldhuizen  S,  Wade  TJ,  Kurdyak  P,  Streiner  DL.  Evaluation  of  2  measures  of   psychological  distress  as  screeners  for  depression  in  the  general  population.  Canadian  journal  of   psychiatry  Revue  canadienne  de  psychiatrie.  2007;52(2):111-­‐20.  Epub  2007/03/23.   9.   Furukawa  TA,  Kessler  RC,  Slade  T,  Andrews  G.  The  performance  of  the  K6  and  K10  screening   scales  for  psychological  distress  in  the  Australian  National  Survey  of  Mental  Health  and  Well-­‐Being.   Psychological  medicine.  2003;33(2):357-­‐62.  Epub  2003/03/08.   10.   Kessler  RC,  Andrews  G,  Colpe  LJ,  Hiripi  E,  Mroczek  DK,  Normand  SL,  et  al.  Short  screening   scales  to  monitor  population  prevalences  and  trends  in  non-­‐specific  psychological  distress.   Psychological  medicine.  2002;32(6):959-­‐76.  Epub  2002/09/07.   11.   Kessler  RC.  The  World  Health  Organization  International  Consortium  in  Psychiatric   Epidemiology  (ICPE):  initial  work  and  future  directions  -­‐-­‐  the  NAPE  Lecture  1998.  Nordic  Association   for  Psychiatric  Epidemiology.  Acta  psychiatrica  Scandinavica.  1999;99(1):2-­‐9.  Epub  1999/03/05.   12.   Meadows  K,  Steen  N,  McColl  E,  Eccles  M,  Shiels  C,  Hewison  J,  et  al.  The  Diabetes  Health   Profile  (DHP):  a  new  instrument  for  assessing  the  psychosocial  profile  of  insulin  requiring  patients-­‐-­‐ development  and  psychometric  evaluation.  Quality  of  life  research  :  an  international  journal  of   quality  of  life  aspects  of  treatment,  care  and  rehabilitation.  1996;5(2):242-­‐54.  Epub  1996/04/01.   13.   Meadows  KA,  Abrams  C,  Sandbaek  A.  Adaptation  of  the  Diabetes  Health  Profile  (DHP-­‐1)  for   use  with  patients  with  Type  2  diabetes  mellitus:  psychometric  evaluation  and  cross-­‐cultural   comparison.  Diabetic  medicine  :  a  journal  of  the  British  Diabetic  Association.  2000;17(8):572-­‐80.   Epub  2000/11/10.   14.   Meadows  K.  Scoring  the  DHP-­‐18.    DHP  Research  &  Consultancy.    113  Lower  Camden,   Chislehurst,  Kent  BR7     5JD  2010.   15.   Tan  L,  Khoo  E,  Griva  K,  Lee  Y,  Amir  M,  Zuniga  Y,  et  al.  Reliability,  Validity  and  Sensitivity  of   the  Diabetes  Health  Profile-­‐18  in  Singapore.  2014.   16.   Snoek  FJ,  Pouwer  F,  Welch  GW,  Polonsky  WH.  Diabetes-­‐related  emotional  distress  in  Dutch   and  U.S.  diabetic  patients:  cross-­‐cultural  validity  of  the  problem  areas  in  diabetes  scale.  Diabetes   care.  2000;23(9):1305-­‐9.  Epub  2000/09/08.   17.   Polonsky  WH,  Anderson  BJ,  Lohrer  PA,  Welch  G,  Jacobson  AM,  Aponte  JE,  et  al.  Assessment   of  diabetes-­‐related  distress.  Diabetes  care.  1995;18(6):754-­‐60.  Epub  1995/06/01.   18.   Welch  GW,  Jacobson  AM,  Polonsky  WH.  The  Problem  Areas  in  Diabetes  Scale.  An  evaluation   of  its  clinical  utility.  Diabetes  care.  1997;20(5):760-­‐6.  Epub  1997/05/01.   19.   Rajaram  R.  Validation  of  the  Problem  Areas  in  Diabetes  Questionnaire  among  patients  with   Type  2  Diabetes  Mellitus  in  Singapore:  A  Pilot  Study.  Singapore:  National  University  of  Singapore,   Department  of  Pharmacy,  2012.   20.   Ismail-­‐Beigi  F.  Clinical  practice.  Glycemic  management  of  type  2  diabetes  mellitus.  The  New   England  journal  of  medicine.  2012;366(14):1319-­‐27.  Epub  2012/04/06.   21.   Sherbourne  C,  Kamberg  C.  Measuring  Functioning  and  Well-­‐Being:  The  Medical  Outcomes   Study  Approach.  Durham,  North  Carolina:  Duke  University  Press  1992:183–93.  1992     22.   Meadows  K.  Scoring  the  DHP-­‐18.  2010.   12  

 

23.   Australian  Mental  Health  Oucomes  and  Classification  Network:  Kessler-­‐10  Training  Manual.   Commonwealth  of  Australia  2005:  NSW  Institute  of  Psychiatry;  2005.   24.   Kazis  LE,  Anderson  JJ,  Meenan  RF.  Effect  sizes  for  interpreting  changes  in  health  status.   Medical  care.  1989;27(3  Suppl):S178-­‐89.  Epub  1989/03/01.   25.   Cohen  J.  Statistical  power  analysis  for  the  behavioral  sciences  (2nd  ed.).  Hillsdale,  New   Jersey:  Lawrence  Erlbaum.  1988.   26.   Wyrwich  KW,  Nienaber  NA,  Tierney  WM,  Wolinsky  FD.  Linking  clinical  relevance  and   statistical  significance  in  evaluating  intra-­‐individual  changes  in  health-­‐related  quality  of  life.  Medical   care.  1999;37(5):469-­‐78.  Epub  1999/05/21.   27.   Machin  D,  Campbell  M,  Fayers  P,  Pinol  A.  Sample  Size  Tables  for  Clinical  Studies,  2nd  Edition.   Blackwell  Science.  Malden,  Mass.1997.   28.   Team  RC.  R:  A  language  and  environment  for  statistical  computing.  R  Foundation  for   Statistical  Computing,  Vienna,  Austria.  ISBN  3-­‐900051-­‐07-­‐0,  URL  http://www.R-­‐project.org/.  2013.   29.   Support  ST.  StataCorp.  2011.  Stata  Statistical  Software:  Release  12.  College  Station,  TX:   StataCorp  LP.  2011.   30.   Thomas  MC,  MacIsaac  RJ,  Tsalamandris  C,  Power  D,  Jerums  G.  Unrecognized  anemia  in   patients  with  diabetes:  a  cross-­‐sectional  survey.  Diabetes  care.  2003;26(4):1164-­‐9.  Epub  2003/03/29.   31.   Beurskens  AJ,  de  Vet  HC,  Koke  AJ.  Responsiveness  of  functional  status  in  low  back  pain:  a   comparison  of  different  instruments.  Pain.  1996;65(1):71-­‐6.  Epub  1996/04/01.   32.   Welch  G,  Weinger  K,  Anderson  B,  Polonsky  WH.  Responsiveness  of  the  Problem  Areas  In   Diabetes  (PAID)  questionnaire.  Diabetic  medicine  :  a  journal  of  the  British  Diabetic  Association.   2003;20(1):69-­‐72.  Epub  2003/01/10.   33.   Singapore  Census  of  Population    2000.  Singapore.  Department  of  Statistics  Singapore:  Oct   2001.  Report  No.   34.   Information  Paper  on  Diabetes  in  Singapore.  In:  Board  HP,  editor.  National  Registry  of   Diseases  Office2011.   35.   Krosnick  JA.  Survey  research.  Annual  Review  of  Psychology.  1999;50:537-­‐67.   36.   Jena  AB,  Press  VG,  Arora  VM.  Social  desirability  bias  in  self-­‐rated  presenteeism  among   resident  physicians-­‐-­‐reply.  JAMA  internal  medicine.  2013;173(2):166.  Epub  2013/01/30.        

 

13  

 

Figure  1:  Graphical  representation  of  the  study  recruitment  of  Singapore  patients  with  T2DM  

   

 

14  

 

Table  2:  Demographic,  Socioeconomic,  Clinical  and  Social  Functioning  characteristics  of  patients  with   Type  2  Diabetes  Mellitus     n=208      

   

   

45.48  

11.87  

   

   

   

Gender  (n,  %)  

   

   

Male  

132  

63.46  

Female  

76  

34.54  

   

   

   

Ethnicity  (n,  %)  

   

   

Chinese  

104  

50.00  

Malay  

24  

11.54  

Indian  

59  

28.37  

Others  

21  

10.10  

   

   

   

Education  (n,  %)  

   

   

<  7  yrs  

16  

7.69  

7-­‐10  yrs  

67  

32.21  

>  10  yrs  

110  

52.88  

Missing  

15  

7.21  

   

   

   

Marital  status  (n,  %)  

   

   

Single  

41  

19.71  

Married  

131  

62.98  

Age,  in  years  (mean,  SD)  

15  

 

Divorced/Widowed  

20  

9.62  

Missing  

16  

7.69  

   

   

    Co-­‐morbidities  (n,  %)  

   

   

Retinopathy    

28  

13.46  

Cardiovascular  Disease  

27  

12.98  

Nephropathy  

17  

8.17  

Neuropathy  

14  

6.73  

Cerebrovascular  Disease  

12  

5.77  

Anemia  

13  

6.25  

PVD  

6  

2.88  

Hepatic  

5  

2.40  

Renal  

1  

0.48  

   

   

   

Poor  Control  (n,  %)  

   

   

No:  HbA1c  (≤7.0)  

60  

28.85  

Yes:  HbA1c  (>7.0)  

148  

71.15  

   

   

   

Presenteeism  (n,  %)  

   

   

Yes  

172  

87.76  

No  

24  

12.24  

   

   

   

Effectiveness  outside  work  (n,  %)  

   

   

Yes  

148  

74.37  

No  

51  

25.63  

16  

 

   

   

   

Family  Functioning  Measure  (n,  %)  

   

   

Good  

184  

89.76  

Poor  

21  

10.24  

   

   

   

Psychological  Distress  Scales  (mean,  SD)  

   

   

DHP-­‐PD  

21.26  

22.02  

K10  

23.71  

17.60  

PAID  

28.73  

21.73  

 

 

 

17  

 

Table  3:  Comparison  of  DHP's  psychological  distress  domain  with  K10  and  PAID               Physical  determinants       Complications   None   Retinopathy     Cardiovascular  Disease   Nephropathy   Neuropathy   Cerebrovascular  Disease   Anemia       Control  (hba1c)   Yes   No       Medication  type   Oral   Insulin   Oral  &  insulin       Social  determinants   Age   <  45  yrs   >=45  yrs         Gender   Male   Female  

    n                   59   28   27   17   14   12   13           60   148           122   7   75               87   120           132   76  

Mean                   19.21   22.02   29.42   34.31   37.70   25.00   17.52           13.80   24.29           17.12   26.98   28.15               22.86   20.09           20.79   22.08  

DHP-­‐PDa   SD   ESd                                   21.84       18.61   -­‐0.13   25.77   -­‐0.44   23.59   -­‐0.68   21.15   -­‐0.85   23.03   -­‐0.26   17.55   0.08                   15.66       23.50   -­‐0.49                   19.63   0.51   19.88   0.05   24.51                               23.52       20.99   0.13                   22.78       20.75   -­‐0.06  

pe                       0.27   0.04   0.01   0.00   0.21   0.62               0.00           1.00   0.56                       0.81               0.34  

                                                                                                                           

Mean                   23.56   26.25   28.70   36.18   35.54   26.25   30.58           18.29   25.91           21.52   21.43   27.27               26.09   21.96           21.27   27.96  

K10b   SD   ES                                   18.27       22.19   -­‐0.14   20.03   -­‐0.27   15.31   -­‐0.71   17.71   -­‐0.66   16.67   -­‐0.15   19.04   -­‐0.38                   15.84       17.84   -­‐0.44                   17.33   0.32   10.49   0.33   18.40                               17.06       17.91   0.24                   16.75       18.32   -­‐0.39  

p                       0.29   0.13   0.00   0.01   0.31   0.11               0.00           0.98   0.90                       0.95               0.00  

                                                                                                                           

Mean                   25.64   31.94   38.15   43.56   39.91   36.88   24.85           20.19   32.19           26.43   23.93   32.86               31.15   26.51           28.16   29.71  

PAIDc   SD   ES                                   20.70       20.24   -­‐0.31   24.95   -­‐0.57   20.86   -­‐0.86   21.36   -­‐0.69   21.78   -­‐0.54   15.85   0.04                   19.52       21.69   -­‐0.57                   21.77   0.30   18.78   0.43   20.90                               21.22       21.48   0.22                   21.06       22.96   -­‐0.07  

p                       0.09   0.01   0.00   0.01   0.05   0.56               0.00           0.98   0.88                       0.94               0.32   18  

 

    Ethnicity   Chinese   Malay   Indian   Others       Marital  status   Single   Married   Divorced/Widowed       Socio-­‐Economic  Status       Education   <  7  yrs   7-­‐10  yrs   >  10  yrs       Housing  type   HDB  1-­‐4  rm  flat   HDB  5  rm/executive  maisone   Private  housing       Household  income   Low  (