The Validity of Morbid Maternal Conditions Captured In Hospital Discharge Data
Abstract Number: S-2
Abstract Type: Original Research
Introduction: The U.S. has among the highest rates of maternal morbidity and mortality in the developed world; this has motivated an increase in surveillance and research efforts aimed at understanding the cause [1,2]. Because of the rarity with which these events occur, efforts have largely relied on hospital discharge data collected primarily for billing purposes.
Little is known about the diagnostic accuracy of the International Classification of Diseases, 9th revision codes used in these studies. We therefore studied the positive predictive value (PPV) of the codes commonly used in studies of severe maternal morbidity.
Methods: We queried the billing records of patients delivered at the Massachusetts General Hospital from 2001 to 2011 to identify those with ICD 9 codes suggestive of maternal morbidity [1,2]. Codes were grouped in 1 of 19 conditions and for each of these conditions, 20 patient charts were selected at random and reviewed. For conditions with fewer than 20 patients, all cases were included.
Pre-specified clinical criteria were established for each condition based on a literature/guidelines review; fulfillment of these criteria served as the “gold standard” for the existence of the condition. Each chart was reviewed by two physicians. When there was disagreement regarding the presence of a condition (n=25), a 3rd reviewer rendered the definitive opinion. The PPV of the ICD9 codes were then determined for each category. The effect of applying additional criteria (i.e. length of stay or ICU admission) on the PPV was also explored.
Results: 235 patient charts were reviewed. Table 1 summarizes the results. For 9 of the conditions, PPV was 100%. For 7 additional conditions, PPV was greater than 75%. Acute heart failure, severe anesthesia complications, and eclampsia were coded less accurately (PPVs ranged from 41.2 to 50%). Imposing a length of stay requirement did not substantially improve PPV for most conditions. Requiring ICU admission increased PPV to 100% for all but 4 conditions.
Conclusion: For most conditions, ICD 9 codes accurately captured morbid maternal conditions during delivery hospitalization, suggesting that billing data can reasonably be used for surveillance and research efforts. Further work is needed to define the sensitivity of the ICD 9 codes and to determine whether the accuracy observed at our institution is generalizable.
1. Obstet Gynecol. 2009;113(2 Pt 1):293-9., 2. Anesthesiology. 2011;115