Table 2.

Algorithm characteristics for the identification of thyroid cancer. Optum Research Database, February 1, 2010 - December 31, 2012

		Adjudicated Status
Algorithm components identified from claims	Total	Cases	Non-Cases	Positive Predictive Value (PPV)	95% CI	Sensitivitya	95% CI	Specificityb	95% CI
a) ≥ 1 claim with ICD-9 193	141	72	69	51%	43-60	100%	94-100	0%	0-7
b) Any inpatient or outpatient claim with ICD-9 193	119	68	51	57%	48-66	94%	86-98	26%	17-38
c) Primary inpatient claim with ICD-9 193	21	12	9	57%	34-77	17%	09-28	87%	76-93
d) Claim for goiter in baseline	49	30	19	61%	46-74	42%	30-54	72%	60-82
e) No dispensings for levothyroxine in baseline	104	62	42	60%	50-69	86%	75-93	39%	28-52
f) No claims with diagnosis of benign thyroid nodule ≤ 60 days after ICD-9 193	123	69	54	56%	47-65	96%	87-99	22%	13-34
g) Claims for thyroid surgery in follow-up	103	70	33	68%	58-77	97%	89-100	52%	40-64
h) Claims for any non-surgical treatment (chemo, radio-iodine, radiation)	40	38	2	95%	82-99	53%	41-65	97%	89-99
i) Any treatment (non-surgical or thyroid surgery)	104	71	33	68%	58-77	99%	91-100	52%	40-64
Algorithms
1) Thyroid surgery and no baseline levothyroxine	90	60	30	67%	56-72	83%	72-93	57%	44-68
2) Thyroid surgery and ≥2 ICD-9 193 codes ≤ 90 days after surgery	68	62	6	91%	81-96	86%	75-93	91%	81-96

CI Confidence interval, ICD-9 International Classification of Disease, 9th Revision

^aThis proxy for sensitivity represents proportion of adjudication-confirmed cases who have the characteristic

^bThis proxy for specificity represents proportion of adjudication confirmed non-cases who do not have the characteristic