In urban and rural India, a standardized patient study showed low levels of provider training and huge quality gaps

Abstract

This article reports on the quality of care delivered by private and public providers of primary health care services in rural and urban India. To measure quality, the study used standardized patients recruited from the local community and trained to present consistent cases of illness to providers. We found low overall levels of medical training among health care providers; in rural Madhya Pradesh, for example, 67 percent of health care providers who were sampled reported no medical qualifications at all. What's more, we found only small differences between trained and untrained doctors in such areas as adherence to clinical checklists. Correct diagnoses were rare, incorrect treatments were widely prescribed, and adherence to clinical checklists was higher in private than in public clinics. Our results suggest an urgent need to measure the quality of health care services systematically and to improve the quality of medical education and continuing education programs, among other policy changes.

Exhibit 3. Adherence To Checklist Of Questions And Exams For Unstable Angina In Madhya Pradesh

source Authors’ analysis. notes Mean values and 95% confidence intervals are shown. All items listed are recommended; those marked “[E]” are essential. A temperature attempt refers to checking temperature either by touch or with a thermometer. EKG refers to either an electrocardiogram performed in the clinic itself or a referral for an electrocardiogram. Beedi-cigarette indicates whether the doctor asked about tobacco use; a beedi is an Indian cigarette consisting of tobacco wrapped in a leaf. “Pain start time” is asked to ascertain a specific time of day.

Exhibit 4. Marginal Effects On Process Quality, Diagnosis, And Treatment Of Provider And Clinic Characteristics In Madhya Pradesh

source Authors’ analysis. notes This exhibit graphs the regression coefficients of three separate regressions of provider and clinic characteristics (an indicator for working in the private sector, an indicator for having no medical qualifications, a facilities and equipment index, and the patient caseload at the time of the standardized patients’ visits) on the percentage of recommended questions asked and exams performed, the relative risk of giving any diagnosis, and the relative risk of giving a correct treatment. Thus, the effects in regression 1 are percentage points. The effects in regressions 2 and 3 are rates—for example, private providers are 1.76 times as likely as public providers to provide any diagnosis. Bars labeled with an asterisk denote statistical significance at least at the 90 percent level of confidence. The other coefficients cannot be significantly distinguished from 0 (for regression 1) or from 1 (for regressions 2 and 3). Each regression also controls for the provider’s sex and experience and contains indicator variables for each standardized patient and for the village’s being in the bottom 20 percent of a wealth distribution implied by an asset index we created from the average household’s ownership of seventeen assets, such as a television or a floor that is not made out of mud. *p < 0.10

Exhibit 5. Marginal Effects On Process Quality, Diagnosis, And Treatment Of Provider And Clinic Characteristics In Delhi

source Authors’ analysis. notes This figure graphs the regression coefficients of three separate regressions of provider and clinic characteristics (an indicator for working in the private sector, an indicator for having no medical qualifications, and the patient caseload at the time of the standardized patients’ visits) on the percentage of recommended questions asked and exams performed, the relative risk of giving any diagnosis, and the relative risk of giving a correct treatment. Thus, the effects in regression 1 are percentage points. The effects in regressions 2 and 3 are rates—for example, private providers are three times as likely as public providers to provide any diagnosis. Bars labeled with an asterisk denote statistical significance at least at the 90 percent level of confidence. The other coefficients cannot be significantly distinguished from 0 (for regression 1) or from 1 (for regressions 2 and 3). Each regression also contains indicator variables for each standardized patient and for the neighborhood’s being in the bottom 20 percent of a wealth distribution implied by an asset index we created from the average household’s ownership of seventeen assets, such as a television or a floor that is not made out of mud. *p < 0.10