Exploring the Use of Diagnostic Codes as an Alternative Approach to Lyme Disease Surveillance in Maryland

BACKGROUND: Lyme disease (LD) is a major public health problem in Maryland, though likely underreported.  Positive LD laboratory results are reported to Maryland local health departments who investigate and classify reports as Confirmed, Probable, Suspect or “Not a Case” for the state surveillance database. In 2009, 4,768 LD reports were entered in the database; 2,029 (43%) were classified as Confirmed or Probable. To explore an alternative surveillance approach using diagnostic codes, we requested these data from healthcare facilities ordering LD tests and surveyed these facilities about clinic characteristics and coding and reporting practices.

METHODS: We requested International Classification of Diseases, Ninth Revision, Clinical Modification (ICD9) codes for the date a LD test was ordered for a 10% random sample (n=474) of 2009 LD reports by case classification. We determined sensitivity and specificity for individual and groups of codes for Confirmed and Probable LD cases compared to those classified “Not a Case”. We also asked facilities to report their specialties, who codes patient encounters, who reports notifiable diseases and whether the facility uses electronic medical records (EMRs). 

RESULTS:  Diagnosis codes were provided for 426 patients by 262 healthcare facilities; data collection is ongoing.  The LD code, 088.81, was the most commonly reported code for each classification: Confirmed (43%); Probable (29%); Suspect (28%); and, “Not A Case” (24%). The sensitivity and specificity of 088.81 were 39% and 76% respectively for Confirmed and Probable cases. Other commonly reported codes fell into logical groups (joint/arthropathy, skin/rash, insect bites and non-specific symptoms), each with low sensitivity and specificity alone and in combination with 088.81.  Facilities that responded to the facility survey (199) ranged from single provider offices to hospitals and reported having Family Medicine (46%) and Internal Medicine (36%) practitioners most commonly. One hundred fifty-eight (79%) reported that clinicians assign ICD9 codes; 105 (53%) reported using EMRs exclusively and 30 (16%) reported using paper and EMRs. Twenty-seven (14%) facilities indicated they do not report notifiable diseases to any health department.

CONCLUSIONS: As an alternative surveillance method, ICD9 codes are not reliable for detecting LD due to its complex clinical manifestations and need for laboratory confirmation.  As medical practices convert to the next version of coding (ICD10), the sensitivity and specificity of LD coding may improve.  In the meantime, alternative and more efficient approaches should be considered, including sentinel surveillance using Family and Internal Medicine practitioners.