Capturing Intravenous Thrombolysis for Acute Stroke at the ICD-9 to ICD-10 Transition: Case Volume Discontinuity in the United States National Inpatient Sample.

Background Transition from International Classification of Diseases (ICD) Ninth and Tenth Revisions (ICD-9 and ICD-10) for hospital discharge data was mandated for US hospitals on October 1, 2015. We examined the volume of patients receiving thrombolysis in ischemic stroke (IS) identified using ICD codes within this transition period in the 2015 to 2016 National Inpatient Sample, a weighted 20% sample of all inpatient US hospital discharges. Methods and Results During the ICD-10 period, 2 case identification strategies were used. Codes for IS were combined with: (1) only the ICD-10 code for thrombolytic given into a peripheral vein and (2) all new ICD-10 codes mapped to the ICD-9 code for all thrombolysis. On visual inspection there was an obvious discontinuity in the volume of patients with IS treated with IV thrombolysis corresponding to 3 time periods: ICD-9 (study period 1), transition (period 2), and ICD-10 (period 3). With Strategy 1, analysis using a linear spline with 2 knots shows that the volume of patients with IS treated with IV thrombolysis was significantly different between study periods 1 and 2 (slope difference -1880, 95% CI -2834 to -928, P=0.005), and periods 2 to 3 (slope difference 1980, 95% CI 1207-2754, P = 0.002). With Strategy 2, volumes did not change significantly between periods 1 to 2, though there was a significant difference between periods 2 and 3 (slope difference 719, 95% CI 91-1347, P=0.034). Conclusions The significant discontinuity in thrombolysis volumes for IS during the transition period for ICD-9 to ICD-10 coding suggests that more rigorous validation of US administrative data during this time period may be necessary for research, resource planning, and quality assurance.

On October 1, 2015, transition from the use of World Health Organization's International Classification of Diseases, Ninth Revision (ICD‐9) to Tenth Revision (ICD‐10) was mandated for all US hospitals covered by the Health Insurance Portability and Accountability Act.

ICD diagnosis and procedure codes are used for health services research, quality assurance, and remuneration. Previous studies have shown good evidence of validity for both ICD‐9 and ICD‐10 codes in identifying patients with ischemic stroke (IS) in administrative data. A recent study including 17 US electronic databases demonstrated no significant changes to IS incidence at the ICD‐9 to ICD‐10 transition using a forward and backward mapping strategy.

Intravenous thrombolysis (IVT) is the standard of care therapy for eligible patients with acute ischemic stroke. In comparison to codes for IS, ICD procedural codes for thrombolysis are not well validated. However, these codes have been used previously to examine changing trends in acute stroke treatment , and the associated impact of hospital characteristics. , Whereas ICD‐9 coding for thrombolysis did not specify by site or method of delivery of thrombolysis, ICD‐10 introduced 10 sub‐codes for thrombolysis depending on site and route of administration. In this descriptive study, we examined the volume of patients with IS treated with IVT at the time of the US transition from ICD‐9 to ICD‐10.

Methods and Results

We utilized 2015 to 2016 data from the National Inpatient Sample (NIS), a publicly available data set sponsored by the Agency for Healthcare Research and Quality. It provides a weighted 20% sample of all‐payer inpatient US hospital discharges. A waiver of consent for use of this publicly available database was obtained from the University of British Columbia Institutional Review Board. Because of the sensitive nature of the data collected, requests to access the data set from qualified researchers trained in human subject confidentiality protocols may be sent to HCUP Central Distributor at hcup@ahrq.gov. Variable coding, national incidence estimation, and statistical analysis were performed using STATA/IC 15.1 (StataCorp LLC, College Station, TX).

All discharges from January 1, 2015 to December 31, 2016 were included. During the ICD‐9 period (January–September 2015), cases were identified using diagnostic codes for IS combined with the procedure code associated with thrombolysis (Table S1). During the ICD‐10 period (October 2015–December 2016), 2 case identification strategies were used: ICD‐10 codes for IS were combined with: (1) only the ICD‐10 code for thrombolytic given into a peripheral vein through a percutaneous approach (Strategy 1) and (2) all related ICD‐10 codes mapped to previous ICD ‐9 code for thrombolysis using the Center for Medicare and Medicaid Services' General Equivalence Mappings, including injection of thrombolytic agent into a peripheral or central vein, artery, or the heart through percutaneous or open approaches (Strategy 2). The volume of patients receiving thrombolysis and then subsequently transferred to another facility (“drip and ship”) was studied using 2 methods: (1) discharges identified using the strategies above and then documented to be transferred to another facility at the end of the admission (2) discharges with ICD‐9/ICD‐10 codes for acute ischemic stroke and thrombolysis in a different facility within the last 24 hours prior to admission to current facility (Table S1). Method 1 represents the sending hospital and method 2 represents the receiving hospital and these volumes should be similar within sample variability.

In an exploratory analysis, the discharge disposition of patients with IS receiving thrombolysis identified using Strategy 1 and those only identified using Strategy 2 was examined. Similar, an exploratory analysis of the proportion of patients with IS receiving thrombolysis also underdoing mechanical thrombectomy on the same admission (using ICD codes listed in Table S2) was conducted.

National incidence estimates are calculated by applying discharge weights to the 20% sampled discharges adjusting for sampling technique. Trend analysis over time was plotted against discharge quarter.

The volume of patients with IS treated with IVT increased every quarter except at the time of the ICD‐9 to ICD‐10 transition. A clear visual discontinuity at the ICD‐9 to ICD‐10 transition was apparent. The 3 time periods identified with inspection corresponded to periods characterized by ICD‐9 coding, the ICD‐9 to ICD‐10 transition, and ICD‐10 coding, respectively (Figure 1). Linear regression using splines with 2 knots at the 3rd quarter of 2015 and 4th quarter of 2015 was used to demarcate 3 study periods: period 1 (Q1–3 of 2015), period 2 (Q3–4 of 2015), and period 3 (Q4 of 2015 to Q4 of 2016).

Figure 1

US hospital discharges for patients with ischemic stroke treated with thrombolysis.

The volume of hospital discharges for patients with ischemic stroke (IS) treated with thrombolysis identified using International Classification of Diseases (ICD) codes. Strategy 1 uses ICD‐10 codes for thrombolytic given in a peripheral vein; Strategy 2 uses all ICD‐10 codes mapped to previous ICD‐9 code.

The effect of time on volume of patients with IS treated with IVT identified using Strategy 1 changed significantly from periods 1 to 2 (slope difference −1880, 95% CI −2834 to −928, P=0.005) and again from study period 2 to 3 (difference in slope 1980, 95% CI 1207–2754, P=0.002). Using Strategy 2, the change from periods 1 to 2 was not significant (slope difference −603, 95% CI −1378 to 170, P=0.096) but there was a significant change between periods 2 and 3 (slope difference 719, 95% CI 91–1347, P=0.034). A similar discontinuity was seen when “drip and ship” volumes were analyzed using Strategy 1 from period 1 to 2 (slope difference −531, 95% CI −865 to −197, P=0.012) and then again from study period 2 to 3 (difference in slope 703, 95% CI 431–974, P=0.002). Using Strategy 2, there was no significant discontinuity in “drip and ship” volumes at the time of the ICD‐9 to ICD‐10 transition. There was no discontinuity effect seen in the volume of IS or on the volume of thrombolysis patients documented as receiving thrombolysis at another facility within 24 hours during this period. The volume of discharges with diagnosis codes associated with IS, IVT using ICD‐9 and ICD‐10 procedural codes, and patients with IS transferred to another facility after IVT and those who received IVT at a prior facility within 24 hours can be found in Table. The volumes of “drip and ship” patients identified in the ICD‐10 period as receiving thrombolysis at another facility within 24 hours at the receiving hospital is consistently higher than the volume at the sending hospital using either Strategy 1 or 2 but the gap is lower using the more inclusive Strategy 2 than Strategy 1.

Table 1

US Hospital Discharges for Patients With Ischemic Stroke 2015 to 2016

	2015Q1	2015Q2	2015Q3	2015Q4	2016Q1	2016Q2	2016Q3	2016Q4
Total discharges	1 797 650 (8 988 249)	1 785 546 (8 927 729)	1 789 213 (8 946 064)*	1 774 617 (8 873 084)*	1 797 772 (8 988 853)	1 771 304 (8 856 513)	1 783 052 (8 915 253)	1 776 301 (8 881 498)
Thrombolysis Strategy 1 †	6353 (31 765)	6356 (31 780)	6394 (31 970)*	3739 (18 695)*	3906 (19 530)	3981 (19 905)	4254 (21 270)	4286 (21 430)
Thrombolysis Strategy 2 †	6353 (31 765)	6356 (31 780)	6394 (31 970)*	6066 (30 330)*	6174 (30 870)	6112 (30 560)	6391 (31 955)	6640 (33 200)
Ischemic stroke	29 569 (147 845)	29 827 (149 135)	29 748 (148 740)*	29 651 (148 255)*	30 388 (151 940)	30 279 (151 395)	30 131 (150 655)	30 756 (153 780)
IS and thrombolysis Strategy 1 †	2070 (10 350)	2155 (10 775)	2208 (11 040)*	1862 (9310)	2028 (10 140)	2134 (10 670)	2152 (10 760)	2245 (11 225)
IS and thrombolysis Strategy 2 †	2070 (10 350)	2155 (10 775)	2208 (11 040)*	2134 (10 670)*	2286 (11 430)	2372 (11 860)	2399 (11 995)	2538 (12 690)
Difference between strategies	N/A	N/A	N/A*	272 (1360)*	258 (1290)	238 (1190)	247 (1235)	293 (1465)
IS and thrombolysis Strategy 1 † and transferred out	1016 (5080)	1005 (5025)	985 (4925)*	856 (4280)*	905 (4525)	893 (4465)	911 (4555)	947 (4735)
IS and thrombolysis Strategy 2 † and transferred out	1016 (5080)	1005 (5025)	985 (4925)*	987 (4935)*	1038 (5190)	1009 (5045)	1046 (5230)	1093 (5465)
IS and received tPA at another facility <24 h ago	913 (4565)	1049 (5245)	1088 (5440)*	1059 (5295)*	1095 (5475)	1167 (5835)	1159 (5795)	1137 (5685)

Numbers reflect sampled discharges (national estimates are shown in bracket). IS indicates ischemic stroke; N/A, not applicable.

Time of mandated transition from ICD‐9 to ICD‐10 in 2015 for US hospitals.

Strategy 1 uses International Classification of Diseases, Tenth Revision ICD‐10 codes for thrombolytic given in a peripheral vein, Strategy 2 uses all ICD‐10 codes mapped to previous ICD‐9 code.

Within the ICD‐10 period, among patients with IS receiving thrombolysis who were only identified using Strategy 2 and not using Strategy 1, a higher proportion died during the admission (9.13% versus 5.39%, χ2 [1, N=11 681]=28.7, P<0.001) and a lower proportion were discharged home (29.6 versus 37.6%, χ2 [1, N=11 681]=55.86, P<0.001) (Figure 2). There was a higher proportion of patients who underwent mechanical thrombectomy during their admission among patients with IS receiving thrombolysis who were only identified only using Strategy 2 (31.9% versus 8.1%, χ2 [1, N=11 681]=666, P<0.001) (Table S3). Baseline demographics comparing patients identified using the 2 different strategies is included in Table S4.

Figure 2

Discharge disposition of ischemic stroke patients identified using different ICD coding strategies.

 

Discussion

We examined the volume of patients with IS undergoing IVT captured using International Classification of Diseases (ICD) coding within a 20% sample of all US discharges at the time of transition from ICD‐9 to ICD‐10. The significant discontinuity in volumes of thrombolysis at the ICD transition in the context of steady volumes of IS suggest that there may be a significant proportion of cases of IS treated with IVT that were not captured when defined by only the new ICD‐10 code specific to IVT. It is not possible to discern the validity of ICD diagnosis codes by examining trends in administrative data, but our findings highlight the need for further direct validation of ICD‐9 and ICD‐10 procedural codes for thrombolysis in acute ischemic stroke against pharmacy or order entry data. However, the relative preservation of thrombolysis volumes using the more inclusive ICD‐10 coding strategy and closer approximation of the “drip and ship” volumes suggest that other thrombolysis‐related codes may have been inappropriately applied for IVT in patients with IS during this transition period.

Acute ischemic stroke patients coded using thrombolysis codes others than peripheral intravenous administration had worse outcomes with a higher proportion dying in hospital and a lower proportion being discharged home. There was also a higher proportion of patients who underwent mechanical thrombectomy in this group and the worse outcome likely reflects a higher proportion of patients with large vessel occlusions. An additional potential explanation for this worse outcome is the capture of thrombolysis use in another context for patients with acute ischemic stroke, such as true intra‐arterial administration during endovascular therapy or possibly for other indications such as pulmonary embolism or myocardial infarction. The period of 2015 to 2016 immediately follows the 5 large landmark randomized control trials showing benefit of mechanical thrombectomy for large vessel occlusion. Given that approximately one‐third of all thrombolysis use in patient receiving mechanical thrombectomy was coded as receiving thrombolysis other than through percutaneous peripheral intravenous administration, (Table S3) this could reflect either (1) a much higher proportion of adjunct intra‐arterial thrombolysis during this period of increased adoption of endovascular therapy, or (2) increased miscoding of the route of administration of thrombolysis in patients getting endovascular therapy.

Our study has limitations. Firstly, while more granular data by month or date of discharge would allow for better visualization and interrupted time series analysis, this information was not available within the NIS. Second, our analysis includes US hospital discharges only and may not be generalizable to other countries. Our findings suggest that future researchers should be cautious in using ICD‐10 codes to study patients with IS treated with IVT, particularly during the time of the transition between ICD‐9 to ICD‐10 coding in the US, and may wish to consider a more inclusive strategy when coding for thrombolysis exposure. Direct validation of ICD procedural codes for thrombolysis against pharmacy or order entry data is a needed future direction for this work.

Sources of Funding

Field is supported by a Sauder Family/Heart & Stroke Professorship at the University of British Columbia, the Heart and Stroke Foundation of Canada and the Michael Smith Foundation for Health Research.

Disclosures

Field has been provided in‐kind study medication by Bayer Canada, and has received speaker's bureau honorarium from Servier and advisory board honorarium from HLS Therapeutics. The remaining authors have no disclosures to report.

Tables S1–S4

Click here for additional data file.