Routine operational data on government programs lack sexiness, and are generally not trendy with Data Revolutionaries. But unlike censuses and household surveys, routine administrative data are readily available at low cost, cover key populations and service providers, and are generally at the right level of disaggregation for decision-making on payment and service delivery. Despite their potential utility, these data remain an under-appreciated asset for generating evidence and informing policy—a particularly egregious omission given that developing countries can leapfrog old, inefficient approaches for more modern methods to collect and manage data. Verifying receipt of service via biometric ID and beneficiary fingerprint at the point of service? India’s already doing it.
To better make the case for routine data, two questions need to be answered—what exactly can be learned from these data and how difficult are they to use?
In a paper just published in Health Affairs with collaborators from the World Bank and the Government of India, we probed these questions using claims data from India’s National Health Insurance Program, Rashtriya Swasthya Bima Yojana (RSBY). Using the US Medicare program as a comparison, we wondered whether reimbursement claims data that RSBY receives from participating hospitals could be used to study the quality of care provided. The main goal was to see how far we could push on an example dataset of hospital claims from Puri, a district in Orissa state.
Here’s what we learned:
First, RSBY collects a wealth of useful routine data. Although RSBY captures fewer data fields than Medicare, there is a lot of crucial and valuable information, from a unique beneficiary ID, diagnosis and procedure category to patient status at discharge and mortality.
Second, many of these data fields suffer from substantial—but easily fixable—problems that affect the accuracy and utility of the generated data. One example are free-text fields that seem to encourage useless entries, like “patient died” in the mortality summary. Other fields suffer from a lack of detail: more than half of the claims in our data were for “general ward, unspecified” which is not very informative of what services were actually provided.
Even so much can be learned that has implications for management and decision-making. We looked at two issues, variations in length-of-stay for the second most common claim type (vaginal hysterectomy) and patterns of claims across hospitals. We found that more than half of the hysterectomy claims had lengths-of-stay above the limit set forth in the RSBY fee schedule, an outcome that has implications for both quality of care and spending efficiency.
There were also many differences in the volume of claims and patient mix across hospitals. This could indicate a mismatch between demand and the provider network in some areas, raises the question of whether RSBY needs to better account for differences in patient profiles and procedures, and should inspire more detailed analyses of variations, e.g., in clinical quality.
For routine data to be useful for policy, administrative data systems need to be well designed, implemented and maintained. All the recommendations of CGD’s Data Revolution work apply to routine data, and improving existing systems should be relatively cheap. In addition, decision-makers need to understand the potential uses of this dormant treasure and become invested in its upkeep. That includes policy-makers and program operators, but also service providers. Just like their counterparts in the US, Indian hospitals could learn a lot from analyzing their own claims data and use this evidence to act.
Our paper indicated clear limitations on what routine data can offer, but also large potential benefits at low cost—the very definition of sexy here at CGD. Using existing systems more efficiently is not only a value for money approach in and of itself, it’s also key to bringing value for money to the provision of public services, including in health.