70240 SlIpportillg Evidence-based Policies am/lmplemefltatioll USING MULTiPLE SOURCES OF INFORMATlON TO ESTlMATE INDlA' S HEALTH WORKFORCE" Krishna D. Rao' , Aamshi Bhalnagart and Peler BernIan: This /lOle describes holV health workjorce estimales were calculated lIsing data from the CenslIs, Nalional Sample Survey Olganizatioll (NSSO) alld Government sources and were trianglllated to produce a set of "best available estimales ". Reslills i lldicale IIlalthere are subSIGlllia{ differences between officially reporred statistics alld eSlimales from lite Censlis and National Sample Survey Organization (NSSO) bUI. impor'alltly. there is belle!" agreemelll between the Jalter two. The Censlts appears 10 be the best available estimate. The lI ole el1l1merales the advantages and disadvantages of each data source and a/so recommends actions to strenglhell bOlh data sources alld data methods. R outine infonnation on India 's health workforce suffers infornlation on the self-reported occupation of all from significa nt limitations reducing its comprehensiveness individuals in the country. Health workforce esti mates and reliability. State professiona l council reports on the arc based on a sample of enumerated individuals. From numbers in the health workforce arc inaccu rate as a resu lt of each di strict of the country, 20% of the rural and 50% of non-adjustment for health workers leav ing the workforce the urban enumeration blocks (EB) were selected using due 10 death , migration and/or retirement or dou ble systematic sampling. In the II smaller states and un ion counting of workers due to their being registered in more territories « 2 mil lion populati on) all EBs were selected, than one state. Further, all states do not follo w the same making the tota l sampl e size roughly 297 million registe rin g procedure which reduces comparabi lity. ' individuals. The sample estimates were then inflated by a Ce rtain c at ego rie s of h ea lth work ers, s uch as factor of 5 for rura l and 2 for urban to get populati on physiotherapists, medical technic ians and faith healers, are total s. not recorded in government stati stics, making it difficult to 2. Nati ona l Sa mple Survey Orga nisation (NSSO), 2004- estimate the overall size and composition of thc health 05: The 61 " round on 'Employment and Unemploy- workforce. ment' , a nationally representati ve household survey, Our study exam ined different data sources such as the reco rded the se lf-report ed eco nomi c activity of Census of India and nationally representati ve household empl oyed individuals covering 124,680 households and surveys to assess how to develop comprehensive and 602,833 ind ividual s. reliable est imates of the health workforce in India . This note 3. Government of Indi a (Go l): Official estimatcs of describes how health workforce estimates from these registered doctors and nurses were obtained from the sources and the Government were triangulat ed to produce a Med ica l and Nursing Counc ils of India. The Mini stry of set of "preferred esti mates". The merits and de-meri ts of this Health and Family Welfare (MO HFW), Government of approach are also discussed. India, also publi shes infornlation on human resources in the health sector through various periodicals, such as DATA SOURCES USED Hea lth Infonnation of India and the Bull etin of Rural 1. Census of Indi a 200 1: The 2001 Census collected Health Stati sti cs . . Health workers in sufficient numbers. in the right places. and adequately trained, motivated and supported arc the backbone of an effective, equitable. and efficient health care system. Success in creating and sustaining an effective health workforce in India to achieve national health goals will require sound policy and creative and committed implementation. More and better infomlation on human resources for health in India is one clement needed to achieve this. This note summarizes recent and ongoing work in support of India's health work force goals. For the fu ll report, sec Rao, K. 1'1 al ·· llIdia:~ Health Workforce: Size, Composition and Distribution ,. HRH Technical Report #1 at www.hrhindia.org I The Public Health Foundation of Ind ia, New Delhi ; : The World Bank, Washington DC METHODOLOGY FOR DEVELOPING A " BEST health workers estimated by the Census and the NSSO were AVAILABLE" ESTIMATE remarkably close to each other. They suggest that overall Several adjustments were perfonncd to make the NSSO and there were2. 1 million health workers in India in 2005 which Ce nslis estimates comparable. Whil e the Censlis estimates translated into a density of approximately 20 health workers were recorded in March 2001 , the NSSO survey was per 10,000 popu lation (Figure I). This estimate included conducted between Ju ly'04 and June'OS . On the assumption both qua lifi ed and unqualified health workers . If the fi gures that growth in the health workforce follows that of the are adjusted to exclude a proportion of those s elf~ re port e d general population, the Censll s estimates were inflated by health workers who did not report the ir qualifi cations (an 8% to refl ect the percent increase in population between estimate of " unqualifi ed" workers), based on the NSSO data 2000-2005 . on se lf~ re port e d qual ifi cati ons, then the adjusted rati o of health workers to population would be a little above 8 per Health workers were identified in the Census and NSSO 10,000 po pulation .1 samples using the National C lass ification of Occupations (NCO). The Census used the NCO-2004 codes whereas the While the tota ls arc close, counts for specific categories of NSSO used the NCO- 1968; NCO-68 codes were converted health workers differed substantially for the different to NCO-04 with littl e loss of infomlation. To furth er sources of data. The Census and NSSO estimates were improve comparabi lity between the two, certain health similar to each other, except for a llopathic physicians and worker catcgories were e ither split or merged together. For ' others' . The Census numbers were higher for a llopathic example, because the fun ction of nurses and midwifes is physicians, nurses and phannaeists . NSSO estimates were often similar they were merged into a single category. Some grea ter for all others. Ga l estimates were generall y higher of the employed individuals in the NSSO data had mi ss ing than the other two, particularl y for nurse and midwifes and NCO codes. These indi vidual s were recogni sed as health for AYUSH practitioners and especiall y higher if one workers based on their National Industrial Classifications considers a correction for unqualifi ed health workers. An (N1C) codes and the ir educationa l qualifi cations . exception to thi s was the slightly hi gher number of doctors reported by the Census. Th is could be a res ult of the The Census and the N SSO classify worker occupations inclusion of inadequately trained medi cal practiti oners in based on se lf~ re port e d occupation descriptions. Thi s thi s category. procedure of identifying and classifying health workers can over estimate the number of qualified health professionals. Both the Census and the NSSO estimates paint a similar For examp le, indi vidual s with a range of qua lifi cations picture of the composition of th e health workforce, practice as all opathic doctors in India. These include especially concerning the adverse nllfse~ph ys i c ian ratio. sp ec iali s ts, ge n e ral pra c tition e r s, rural me di c al They also suggest a very low proportion of female health practitioners, and others with no formal training or workers, espec ially female physicians and parti cularly so in certifi cation in medic ine. In the NSSO data, 25% of the rural areas. individuals (42% in rural and 15% in urban) classified as Figure 2 compares the state leve l estimates of the number of doctors reported no medical tra ining. Whil e it is not poss ible health workers produced by the Census and NSSO. The to verify how many doctors (or other health workers) maj ori ty of states tend to cluster close to th e diagonal line, identifi ed in the Census were full y qua lifi ed, it is possible indicating good correspondence between the Census and that similar level s of less than full y qua lified medical NSSO estimates. However, there arc significant differences profess ionals arc also included. Thi s would result in a large behveen the two sources of data even when total health overestimate o f the numbers of qual ifi ed phys icians . There worker numbers are summed by state. And there arc some is no representative way of assess ing this prob lem at this large outlie rs (e.g. Gujarat, Sikkim , De lhi and Mi zoram), time. which fa ll at a substantial di stance away from the diagonal The fina l set of health worker categori es for which estimates line showing poor agreement between the NSSO and were produced included ~ allopathic physicians and Census. Disaggregated results for allopathi c doctors and surgeons, dentists, AYUS H practitioners (Ayurvedic , Yoga, nurses and midwifes a lso show a similar pattern . Since the Unani , Sidha, Homeopathy), nurses and midw ives, overall Census and NSSO estimates arc close, these int e r~ pharnlaci sts, others (d ietic ians, opticians, dental assistant s, state di screpancies cancel out each other at the aggregate phys iotherapists, medica l ass istants and technicians and level. State level estimates using these two sources are much other hospital sta ff) and other traditi onal medi cine less consistent than nati onal aggregates. praetitioners. The Census and NSSO estimates for di stribution of health COMPARISON OF DIFFERENT DATA SOURCES workers across rura l and urban areas were a lso similar. Both The Census, NSSO and Gol estimates for the number of sources indicate that overall , and amongst most health health workers overall and in different categories are shown worker categories, typicall y 60% of the health workers in Figure I . A comparabl e Government estimate ohhe total resi ded in urban areas. Government estimates of the number number of health workers does not exist. The number of of hea lth workers in rural and urban areas were onl y P\DI.\ III .\L 111 Iii \ J. \'Illumc I • \iumhcl 2 ( \ug\l"[ 2()()l) I available for those working in the public sector. Figure 1: Health Worker Density - All India, 2005 (Per 10,000 Population) Estimates of heath workforce based on the Census and NSSO data ha ve several ~~~~;;;;;. advantages. They are based on population Allopathic Physician Nurse & Midwife L AYUSH '--'" .. counts in each stale wh ich avo ids the Dentisl problem of double counting. They are Pharmacisl ~ available fo r all states in India , prov ide Others geographi c estima tes, and have fewer Other Traditional ~ All " : : : : : : : : : : : : : : " ' ".................." "......................,. comparabil ity issues because they are based on standard occu pationa l codes. However AII' (excluding unqualified) r thcse sources of data have limitations. The o 2 4 6 8 10 12 14 16 18 20 small samp le sizc of the NSSO prevents _ NSSO _ Census Government robust di saggregaled estimates at the stale • Estimates based on seH·repor1ed occupation In NSSO level and by health worker type. Further, these data sources cannot track sho rt ~ t enn Source: Natlonat Sample SUNey Organisation !NSSO) 2004·05; Census 01 India 2001; Medical and Nursing Couoclls 01 Indie; Government o( India. Central Bureau 01 Hea~lllntelligence changes in the health workforce since the Otllers . DIetician & Nutritionist. Oplicial'ls. Dental Assistant. Pilysiotilerapist. NSSO survey is conduc ted onl y every 5 Medical ASSistant & Technician and Otller Ho~tal SIal! Otller Traditional", Trad~ional Mediclr>e Practitioner. Faith Healer years and the Census every ten years. CONCLUSIONS Seve ra l data so urces pro vi de usab le Figure 2: Comparison of NSSO and Census Estimates of infonnation on the health workforce in India. Health Worker Density' in Different States 2005 W We conclude that, at this time, the Census is o GA the preferred source fo r overall hea lth workforce esti mates. The census has the advantage of being an officially accepted data source for government. Its large sample size provides estimates for every di strict in o H l the country and, withi n eac h di strict, both urban and rural areas. It also supports robust z "' "' estimates of the health workforce ac ross O M' hea lth wor ker ca tegor ies, states and O Dl geograph ica l areas. Further, the Census estimates have becn shown to have good a correspondence with the NSSO cstimates at th e aggregate leve l, indicating so me o 20 40 60 CENSUS reliability. However, both the Census and the • Del'lsity Per 10.000 Population NSSO esti mates are based on sc l f~repo rt ed occupati ons wh ich is susceptible to incorrect ;o.ource: National Sample Survey (NSSOI2004-05 . Census 01 India 2001 se lf~re ports . particularly by unqualified health workers. categories of health workers and state level estimates. Since For breakdowns by public and private sector, on ly the states are the largest employers of hea lth workers and the NSSO data is avai lable fo r nationa l and state levcl main regu lators of the health workforce, effective state- estimation at this ti me. We havc applied its public and level policy making will be serious ly hindered by private sector percentages to the census numbers for this uncertainty about numbers, types of workers, location, and purpose. sector of work. Wh ile we propose use of the avai lable data at this time, th is Urgent action is needed to improve data for pol icy and docs not mean the data are sati sfactory. The obvious plann ing. We propose act ion to strengthen both data sources inconsistencics. such as those regard ing physic ian training and methods. and qualifications, are unacceptable. India shou ld be able to COlint its medical doctors as we ll as other major categories ACTIONS TO IMPROVE OATA SOURCES AND of hea lth workers in a timely, accurate, and reliable way. METHODS Despite similarities in tota l numbers between the censLis and I. Create a task force compri sing of the MOHFW. agencies the NSSO estimates, there are large inconsistencies across coll ecti ng human resource infonnation (e.g. Census, ['\[)[ '\ II I \[ II [ iiI \ I \'Il[UllK' [ . .... ul1lh~·r ~ ( \Ul!u,,1 .::'O()'J) NSSO) and experts to recommend ways to make human reported occupations through better assessment of reso urces information more rel iabl e, val id and timely. education and other cri teria. The upcomi ng Cens us in 2. Task foree should review registration procedures for 20 II st rengthened by modifyi ng its questionnaire in a d oc tors, n urses, m id w ives, an d ph ar mac is ts . manner which wou ld separate out the professionall y Professional co unci ls should be encouraged to maintain trained medical practitioners fro m those who are not. live reg isters and updated on a conti nuous basis. Other Si milarly, the use of new cl assification codes to health workers, especially RM Ps, shou ld also be segregate nurses from midwifes. simi lar! y registered. 4. For data sources like the Census and NSSO, create new 3. For data sources like the Census and NSSO which rely on elassification codes for categories of health workers like sel f-report ed occ upatio ns, improve va lidation of community hea lth workers and RMPs. I Delhi Medical Counci l requires doctors practicing in Delhi to register themselves every five years, a practice which is not followed in other state medical councils. ' There is no fully satisfactory way to adj ust census figures for qualifications at this time, but this adj ustment is presented to emphasize this problem in the available data. Data on health workforce presented in this policy note do not distinguish between qualified and unqualified health workers, unless specifically mentioned. REFEREN CES I. Anand Sand BamighauseTl T. 2004. '·'·Iuman resources and health outcomes: cross-country econometric study". Lancet, 364: 1603- 09. 2. Anand Sand Barnighausen T. 2007. "Health Workers and Vaccination Coverage in Developing Countries: An Econometric Analysis". Lancet, 369: 1277-1285. 3. Banerjee A. Deaton A and Duno E. 2004. Wealth. I-Icalth and Health Services in Rural Rajasthan. Paper No.8. Poverty Action Lab, Massachusetts Institute ofTechnology. 4. Census oflndia. 200 I. http ://www.censusindia.gov.in 5. Government of India. 1961. Report of Hea lth Survey and Planning Comminee (Chainllan: Mudaliar), Ministry of Health and Family Welfare, Government oflndia. 6. Government of India. 2004. National Occupational Classification. hltp :lldget.nic.in/ nco/jobdescription/welcome.html 7. Governmentoflndia. 2005. Central Bureau onlca lth Intel ligence. http://www.cbhidghs.nic.in 8. Governmentoflndia. 2005. I-I uman Resources for Health. In Financing and Deliveryo fl'lcalth Services in India. National Commission on Macroeconomics and Health Background Papers. Ministry of Health and Family Welfare, Government oflndia. 9. Government of India. 2006. Bulletin on Rural Health Statistics in India. lnfrastmcture Division, Department of Family Welfare. Ministry of Health and Fami ly Welfare, Governmem oflndia. 10. International Institute of Population Sciences.2005. National Family Health Survey (NFH S-3), 2005-06, India. International Institute of Population ScicncesandORC Macro: Mumbai. II. Joint Learning Initiative. 2004. ['Iuman Resources for Health - Overcoming the Crisis. Joint Learn ing Initiative, Harvard University and World Health Organization. 12. Medical Counci l of India. 2005. http://www.mciindia.org 13. National Sample Survey Organisation. 2004-05. 61 ~ Survey Round on Employment and Unemployment in India. National Sample Survey Organisation. New Delhi. 14. World Development Repon. 1993. World Bank. Washington DC. 15. World Health Organisation. 2006. Working Together for Health - World l'lealth Repon 2006. World Health Organization. Geneva. 16. World Health Organisation. 2007. Not Enough Here .. Too ManyThere - I'lealth Workforce in India. World ['Iealth Organizat ion. Country Office for India. Editors: Gerard La Forgia. Lead Specialist. HNP Unit. The World Bank. New Delhi ; and Krishna D. Rao. Public Hea lth Foundation 01" India, New Delhi India Heal/Ii Beal is produced by the Public Health Foundation of India and the World Bank's Health, Nutrition and Population unit located in Delhi. 111e Notes arc a vehicle for disseminat ing policy-relevant research. case studies and experiences pertinent to the Indian health system. We welcome submissions fTOm Indian researchers and the donor community. Enquiries should be made to Nira Singh (nsingh2@worldbank.org). Disclaimer: The views, findings, interpretations and conclusions expressed in this policy note arc entirely of the authors and should not be attributed in any manner to the World Bank, its affiliated organizations. members of its Board of Executive Directors. the countries they represent or to the Public Health Foundation of India and its Board of Directors.