Our first aim was to determine the accuracy of the administrative code for ITP, ICD-9-CM 287.3, in children and adults. Data from this study confirmed previous results of Segal and Powe(5
) who reported a PPV of 0.71 (95% confidence interval 0.63, 0.79) for outpatients of all ages managed by all physicians at Johns Hopkins Hospital. Our overall outpatient PPV of 0.65 was not different. Additionally, our data also demonstrated similar PPVs when data were restricted to hematologists and children were distinguished from adults: 0.72 for children and 0.69 for adults. Therefore when a hematologist billed a child for ITP, the child had a 72% chance of having a definite diagnosis of ITP and an adult had a 69% chance of having a definite diagnosis of ITP. While no specific threshold for PPV is available for this type of study, thresholds specified for measures of reliability, such as a Kappa statistic, would classify the measures of 0.72 and 0.69 as substantial.(10
) In reality, thresholds are somewhat specific for the condition under study and for the needs of the study.
Our second aim was to document that hematologists are the principal physicians for patients with ITP and therefore to support the methodology that limiting administrative data to hematologists’ records can identify most patients with ITP. In 98% of children and 92% of adults, the definite diagnosis of ITP was established by a hematologist rather than a non-hematologist. Even including patients with a probable diagnosis of ITP, in 97% of children and 86% of adults the diagnosis was established by a hematologist. These data document that limiting the analysis of administrative data to hematologists’ records can identify most patients with ITP and are consistent with our previous survey of physician referral patterns for patients with thrombocytopenia. This survey of primary care providers in the Oklahoma Practice-based Research Network documented that 75% of respondents were ‘likely’ to send a patient with moderate thrombocytopenia (platelet count 30,000/µL) to a hematologist for further evaluation and management.(11
) The likelihood of referral increased to 85% when the moderate thrombocytopenia was associated with mild bleeding symptoms (petechiae) and to 92% when the patient presented with severe thrombocytopenia (platelet count 10,000/µL).(11
Our third aim was to document that most patients with ITP have at least part of their care as outpatients. Five percent of children and 17% of adults with a definite diagnosis of ITP by a hematologist were only identified by inpatient medical records. However, 7 of the 12 children and 11 of the 34 adults only identified by inpatient medical records had scheduled outpatient follow-up appointments with hematologists. Therefore we determined that limiting medical review to outpatient records would identify 98% of children and 91% of adults with a definite diagnosis of ITP by a hematologist.
Combining the results comparing hematologists to non-hematologists and outpatient records to inpatient records, these data document that 96% of all children and 79% of all adults who have a definite diagnosis of ITP can be identified from the outpatient records of hematologists.
Our data together with previous data(5
) demonstrated that that misclassification was rarely due to diseases other than ITP that were also correctly coded by ICD-9-CM 287.3. Almost all misclassification was due to diseases being incorrectly coded. Seventy-seven children were coded incorrectly as 287.3; many had leukemia. Seventy-seven adults were coded incorrectly; many had thrombotic thrombocytopenic purpura, gestational thrombocytopenia, or essential thrombocythemia. Therefore, even with the refinement of the ICD-9-CM 287.3 code which now has specific codes for each disease other than ITP within that code (October 1, 2005), the frequency of coding inaccuracies may not change.
An important element of this study was to establish levels of evidence for the diagnosis of ITP established by medical record review to identify patients with a definite diagnosis of ITP. To develop these levels of evidence required extensive review of medical records and interpretation of notes written by multiple physicians of different medical specialties. As our experience with medical record review encountered many complex situations, our system evolved to the final levels of evidence reported here. Although some patients designated as “probable” ITP may have had ITP, the number of patients in this category was small, 3% of children and 4% of adults.
A limitation of this study is that it only analyzed patients at the OUMC and the results may not be generalizable to community hematologists throughout Oklahoma or other states or to other administrative databases. The OUMC has all medical specialties within a single medical center, and OUMC patients may be seen by medical specialists more frequently than patients receiving community care. More patients at OUMC may be uninsured and may be more seriously ill than patients cared for by community physicians; this may also result in care by a greater variety of medical specialists. The greater number of physicians of different medical specialties may decrease the role of hematologists. On the other hand, referral among physicians may be greater within the OUMC resulting in a greater fraction of patients with ITP managed by hematologists. With other administrative databases, it may not be feasible to identify data restricted to hematologists. However, the PPV in our study was similar when all patients with a definite diagnosis of ITP were analyzed compared to patients diagnosed by hematologists. However, as expected, ITP was more likely to be the principal diagnosis for hematologists, coded in the first position, than for non-hematologists.
A strength of this methodology is that it may be a model for epidemiologic studies of other uncommon disorders which do not have explicit diagnostic criteria and are primarily managed by specialists as outpatients. Additional strengths of this study are the consistency with the previous data(5
) and the completeness of the medical record review. For children and adults, respectively, 98% and 95% of eligible medical records across 10 years were reviewed.
In conclusion, these data document the validity of a method for identifying patients with ITP from diverse data sources utilizing administrative records, regardless of patients’ age and source of healthcare. Confirmation of the accuracy of the ICD-9-CM code for identifying patients with ITP is important to establish the validity of epidemiologic studies using administrative databases. Limiting analysis of administrative data to hematologists’ outpatient records provided a feasible method to identify most children and adults with ITP.