ABSTRACT
The incidence of hypertensive (HTN) crises in older adults has more than doubled over the past two decades. ICD‐10 introduced codes for HTN urgency and emergency, but their validity has not been evaluated. We analyzed a 20% national sample of fee‐for‐service Medicare beneficiaries aged ≥66 years (2016–2019). Medical records were reviewed for 220 randomly sampled emergency department visits or hospitalizations with a primary ICD‐10 HTN crisis code (100 HTN urgency [I16.0], 100 HTN emergency [I16.1] and 20 unspecified [I16.9]). Physicians abstracted and adjudicated cases. Two gold standards (GS) were used: (1) treating physician diagnosis and (2) adjudicated diagnosis based on BP (SBP ≥180 mmHg or DBP ≥110 mmHg), with or without end‐organ damage. PPVs and 95% confidence intervals were estimated. Among 220 charts sampled from 53,100 HTN crisis cases (mean age 73 years; 33% male; 66% White), PPV for HTN emergency (I16.1) was 91% (95% CI, 85%–96%) using the treating physician GS and 78% (95% CI, 71%–86%) using the adjudicated GS. PPV for HTN urgency (I16.0) was 97% (95% CI, 93%–99%) and 81% (95% CI, 72%–88%), respectively. Overall PPV for any HTN crisis (I16.X) using the adjudicated GS was 95% (95% CI, 91%–97%). Including unspecified codes reduced PPV to 86% (95% CI, 79%–92%) and 78% (95% CI, 71–86%). ICD‐10 HTN crisis codes demonstrate high PPV, supporting their use in epidemiologic and health services research.
1. Introduction
Uncontrolled hypertension (HTN) remains a significant public health challenge, as less than half of the 115 million U.S. adults with HTN have achieved optimal blood pressure (BP) control [1, 2]. Furthermore, the incidence of hospitalizations for HTN crises has doubled to nearly 125/100,000 person‐years over the last 20 years [3, 4], contributing to $2.8 billion in direct medical costs [5].
Historically, studying HTN crises (i.e., HTN urgency and emergency) in large population‐based data sources such as Medicare was not straightforward because the International Classification of Diseases (ICD‐9) coding system lacked dedicated codes. Previous studies that studied HTN crises in large healthcare utilization data using ICD‐9 codes employed various expert‐driven yet unvalidated definitions for HTN crises [4, 6, 7]. The transition to the ICD‐10 system that offers designated codes for HTN urgency and emergency has provided an opportunity to identify HTN crises more accurately. However, validating these ICD‐10 codes in a national sample is necessary for research use, particularly considering heterogeneous adoption patterns and variability in physician coding behavior [8].
Thus, we conducted a study to determine the positive predictive value (PPV) of the ICD‐10‐based HTN crisis codes compared to multiple gold standards derived from medical record review in a national sample of Medicare beneficiaries.
2. Methods
2.1. Data Sources and Study Population
Using a 20% random sample of fee‐for‐service (FFS) Medicare beneficiaries who were eligible for Parts A, B, and D between January 2016 and December 2019, we identified inpatient hospitalizations and/or ED visits among beneficiaries aged ≥66 with ICD‐10 codes for HTN crises.
2.2. Claims‐Based Definitions for Hypertensive Crisis
HTN emergency or urgency visits were defined as inpatient hospitalizations or ED visits (not resulting in inpatient hospitalization) with an ICD‐10 diagnosis of I16.X in the primary diagnosis position. For HTN emergency, we included ED visits resulting in inpatient hospitalization. For HTN urgency, we included ED‐only visits as well as ED visits resulting in inpatient hospitalization. We included ED‐only visits for HTN urgency because such cases usually do not require acute management and can be discharged with outpatient follow‐up.
2.3. Medicare Record Retrieval
Of the 53,100 HTN urgency or emergency cases identified from Medicare data, we used National Provider Identifier (NPI) numbers to link these medical records with hospital addresses and contact information for 41% (N = 22,000) of cases. The CMS identified the social security numbers of the patients for these hospitalizations, which were then sent securely to a third‐party clinical research organization, Information Collection Enterprises, (ICE), York, PA. ICE sent requests for a random sample of 660 records to reach a target number of 220 records. The sample size was determined based on the expected PPV of 85% with a 95% confidence interval lower bound being 80%, yielding a minimum sample size of 50 records for HTN crisis or emergency [8].
We opted to collect 100 records for each of the HTN emergency and urgency ICD‐10 codes to characterize the sample more precisely. We additionally collected 20 records to explore the validity of ‘unspecified HTN crisis’ code. Ultimately, we obtained 100 records with a primary diagnosis code of ICD‐10 code I16.1 (HTN emergency), 20 records with a primary code of I16.9 (HTN crisis, unspecified), 80 records of ED‐only visits with I16.0 (HTN urgency), and 20 records of inpatient hospitalizations with a ICD 10 code of I16.0 (HTN urgency) as the primary diagnosis (Figure 1).
FIGURE 1.
Validation sample retrieval cascade. This figure shows the cascade to identify patients with hypertensive crises ICD‐10 codes in a 20% Medicare Parts A, B, D sample with complete NPI information. The hospital contact information is then used to retrieve charts via a third‐party adjudicator.
2.4. Gold Standard Definitions for Hypertensive Crises in Medical Records
We developed multiple gold standard (GS) definitions for HTN crisis based on the information available in medical records. For HTN emergency, the primary GS of HTN emergency was a documented diagnosis by a treating physician in the discharge summary (Physician Diagnosis). The secondary GS was based on adjudication by trained physicians based on documented admission SBP ≥180 mmHg or DBP ≥110 mmHg AND evidence of end‐organ damage (EOD) in the medical record (Adjudicated Diagnosis). We used DBP ≥110 mmHg rather than 120 mmHg as noted in the 2017 and 2025 ACC/AHA guidelines, as our study period included years prior to publication and adoption of new thresholds [8]. EOD was defined as any of the following noted in medical records: (1) neurological: new neurological deficits, altered mental status, CT/MRI findings of acute ischemic or hemorrhagic stroke, other intracranial hemorrhage, increased intracranial pressure, head trauma; vision changes, evidence of papilledema, retinal hemorrhages, grade III/IV hypertensive retinopathy; (2) cardiac: new EKG changes, structural echocardiogram changes, elevated troponins, diagnosis of acute coronary syndrome (unstable angina, non‐ST elevation myocardial infarction [NSTEMI], STEMI); (3) renal: acute kidney injury (serum creatinine (Cr) increase by ≥0.3 or 50% increase in serum Cr compared to serum Cr before last non‐HTN crisis BP reading), acute renal failure, anuria/oliguria; (4) vascular: CT or MRI angiography evidence of aortic dissection, Chest X‐ray (CXR)/CT evidence of widened mediastinum, double aortic contour; (5) pulmonary: flash pulmonary edema on CXR/CT [5, 6, 9]. We also used an additional GS, termed ‘Adjudicated Diagnosis + Critical Care’, which comprised meeting criteria for the adjudicated GS + admission to a critical care unit or use of IV BP‐lowering medication in order to more specifically identify HTN emergency which typically is recommended to be treated in an intensive care unit.
HTN urgency, more recently renamed to severe asymptomatic HTN, employed a GS definition of treating physician diagnosis (Physician Diagnosis) as well as an adjudicated GS of admission SBP of ≥180 mmHg or DBP of ≥110 mmHg without a record of EOD on admission (Adjudicated Diagnosis). As in HTN emergency GS criteria, we chose 110 mmHg as the cutoff for DBP because the 2017 revised ACC/AHA HTN guidelines that increased the threshold to 120 mmHg were not in effect in 2016 and likely not adopted in 2017.
The GS for HTN crisis (either HTN emergency or urgency) was defined using the treating physician's diagnosis for either HTN urgency or emergency (Physician Diagnosis) as well as the adjudicated diagnosis based on recorded BP values on admission (SBP ≥180 mmHg or DBP ≥110 mmHg) (Adjudicated Diagnosis).
2.5. Data Abstraction by Trained Physicians
Study data were collected and managed using REDCap (Research Electronic Data Capture) tools hosted at Rutgers University. REDCap is a secure, web‐based software platform designed to support data collection for research studies [9]. We developed a medical record abstraction form to systematically collect and store information from a patient's medical record related to the HTN crisis hospitalization, including symptoms, physical examination, laboratories, imaging, and BP medication use and self‐reported adherence. Given the need to manually abstract information from records and occasional ambiguity associated with differentiating HTN urgency from emergency, we trained four physicians‐in‐training for data collection. Each trained physician abstracted information from the patient's medical record into the REDCap form. Abstractors were blinded to information about the patient's corresponding ICD‐10 codes in Medicare data. The abstraction form is included in the supplement.
Inter‐Rater Reliability: Given the need to manually abstract information from records and occasional ambiguity associated with differentiating HTN urgency from emergency, we trained four physicians‐in‐training to abstract data from records and assessed the abstractors’ inter‐rater reliability. Amongst the six records (out of 55 records abstracted by each physician) assessed for adjudicator agreement, all four adjudicators agreed 83% of the time based on the adjudicator's interpretation of HTN emergency hospitalization, 67% of the time for diagnosis by treating physician, 67% for BP≥180/110 and evidence of EOD, and 100% of the time for HTN urgency (Table S1 in the supplement). To reconcile these differences, three separate physicians additionally verified adjudication on 66–67 records each and settled discrepancies based on the data abstracted.
2.6. Statistical Analyses
We first described the characteristics of the 220 patients in our validation sample as compared to HTN crises cases in a 20% random sample of Medicare‐FFS data by age, demographics, and HTN‐related diagnoses. We then described the characteristics of the validation sample by HTN urgency and emergency. One‐way ANOVA was used to compare continuous variables and X2 for categorical variables, assuming normality. The PPVs and 95% CIs were calculated for each GS by calculating the proportion of patients with true HTN urgency or emergency as per GS amongst those with ICD‐10 codes for HTN urgency (I16.0) or HTN emergency (I16.1 alone). We also calculated the PPV for detecting any HTN crisis hospitalizations (I16.X, either HTN emergency or urgency) against a GS of physician and adjudicated diagnoses. As a sensitivity analysis, we determined the PPV for HTN emergency using both HTN emergency and unspecified HTN crisis ICD‐10 codes (I16.1 or 16.9). All analyses were performed with SAS version 9.4 (SAS Institute, Cary, NC). PPVs are presented as PPV % (95% CI). The study protocol was approved by the institutional review board at Rutgers University.
3. Results
Our sample included 220 cases with medical records randomly selected for retrieval from all (N = 53,100) HTN emergency or urgency cases (mean age 73y, 33% male, 66% White). They were not significantly different with regards to demographics or comorbidities as compared to a national, random sample of 22,000 Medicare beneficiaries above 65 years (Table 1 ). Both the study sample and national sample had approximately 97% with HTN, 45% CKD, 50% with diabetes, 40% with heart failure, and 48% with ischemic heart disease. Comparisons within each ICD‐10 code group are presented in Table S2. Among the HTN urgency group of 20 inpatient hospitalizations, age was significantly different between the national and study samples (74y vs. 80y, p = 0.006). No other significant demographic or comorbid differences by ICD‐10 code group were observed.
TABLE 1.
Comparison of demographic and comorbid factors in the validation vs. random Medicare sample.
| Demographics | Random sample (n = 22,000) | Validation Sample (N = 220) | p values |
|---|---|---|---|
| Mean Age (SD) | 72.1 (14.4) | 73.1 (13.4) | 0.15 |
| Male, % | 7083 (32%) | 72 (33%) | 0.87 |
| Race/ethnicity | |||
| White | 14,376 (65%) | 145 (66%) | 0.87 |
| Black | 5663 (26%) | 61 (28%) | |
| Hispanic | 300 (1%) | 6 (3%) | |
| Asian | 518 (2%) | 4 (2%) | |
| North American Native | 792 (4%) | 1 (1%) | |
| Other | 137 (1%) | 0 (N/A) | |
| Unknown | 214 (1%) | 3 (1%) | |
| Comorbidity | |||
| Substance use disorder | 2203 (10%) | 17 (8%) | 0.26 |
| Diabetes mellitus | 10,819 (49%) | 109 (50%) | 0.91 |
| Diabetes w/ complications | 7850 (36%) | 76 (35%) | 0.73 |
| Hypertension | 21,399 (97%) | 215 (98%) | 0.68 |
| Chronic kidney disease | 9635 (44%) | 101 (46%) | 0.53 |
| End‐stage renal disease (ESRD) | 2826 (13%) | 32 (15%) | 0.46 |
| Heart failure | 8705 (40%) | 88 (40%) | 0.70 |
| Stroke | 6363 (29%) | 55 (25%) | 0.20 |
| Ischemic heart disease | 10,560 (48%) | 105 (48%) | 0.94 |
Among the 100 patients with an ICD‐10 claims diagnosis of HTN emergency (I16.1), the mean age was 71 years (SD, 14) and 42% were male. The mean SBP and DBP were 217 (SD, 23) and 105 (SD, 22). Nearly 98% of patients had a prior diagnosis of HTN, 92% were prescribed antihypertensive medications, 57% were taking 3 or more antihypertensive medications of which beta blockers (62%) and ACEi/ARBs (57%) were most common, and 15% had prior hospitalizations for HTN crisis (Table 2). The most common presenting symptoms were headache (32%), dizziness (25%), chest pain (32%), or shortness of breath (41%). Median serum creatinine (IQR) on admission was 1.4 (1.1–5.0). Overall, 41% eventually were admitted to a critical care unit, 70% were administered an IV push antihypertensive medication, and 50% were started on IV drip medications. Common objective measures of acute or chronic EOD included left ventricular hypertrophy (31%) and pulmonary edema (26%).
TABLE 2.
Baseline characteristics in hypertensive urgency and emergency Medicare beneficiaries.
| Characteristic |
HTN crisis (I16.X) N = 220 |
HTN urgency (I16.0) N = 100 |
HTN emergency (I16.1) N = 100 |
HTN crisis, unspecified (I16.9) N = 20 |
|---|---|---|---|---|
| Demographics via claims | ||||
| Age (SD) | 73.4 (14.6) | 75.4 (12.2) | 70.6 (14.1) | 74.7 (14.8) |
| Female (%) | 148 (67%) | 76(76%) | 58 (58%) | 13 (65%) |
| Race | ||||
| Asian | 4 (2%) | 3 (3%) | 0 (N/A) | 1 (5%) |
| Hispanic | 6 (3%) | 3 (3%) | 3 (3%) | 0 (N/A) |
| Black/African American | 61 (28%) | 20 (20%) | 36 (36%) | 5 (25%) |
| White | 145 (66%) | 73 (73%) | 59 (59%) | 13 (65%) |
| Other/Unknown | 4 (1%) | 1 (1%) | 2 (2%) | 1 (5%) |
| Care settings | ||||
| ED visit only | 62 (28%) | 62 (62%) | N/A | N/A |
| ED visit + Inpatient hospitalization | 113 (51%) | 35 (35%) | 61 (61%) | 17 (85%) |
| ED visit + Inpatient + ICU admission | 45 (20%) | 3 (3%) | 39 (39%) | 3 (15%) |
| Presenting symptoms | ||||
| Headache/migraine | 79 (36%) | 44 (44%) | 32 (32%) | 3 (15%) |
| Alertness and orientation | ||||
| Alert and oriented x 1 | 7 (3% | 1 (1%) | 6 (6%) | 0 |
| Alert and oriented x 2 | 5 (2%) | 2 (2%) | 3(3%) | 0 |
| Alert and oriented x 3 | 177 (80%) | 81 (81%) | 83 (83%) | 13 (65%) |
| Not documented | 31 (15%) | 16 (16%) | 6 (6%) | 7 (35%) |
| Altered mental status (e.g., stupor, coma) | 32 (15%) | 4 (4%) | 22 (22%) | 6 (30%) |
| Dizziness/vertigo/unsteadiness | 53 (24%) | 26 (26%) | 25 (25%) | 2 (10%) |
| Chest pain/chest discomfort | 60 (27%) | 20 (20%) | 32 (32%) | 8 (40%) |
| Shortness of breath/dyspnea | 63 (29%) | 12 (12%) | 41 (41%) | 10 (50%) |
| Focal neurological deficit | 12 (5%) | 3 (3%) | 8 (8%) | 1 (5%) |
| Antihypertensive use at home | 201 (91%) | 91 (91%) | 110 (92%) | 17 (85%) |
| Number of antihypertensive classes | ||||
| 0 | 19 (9%) | 9 (9%) | 7 (7%) | 3 (15%) |
| 1 | 39 (18%) | 22 (22%) | 14 (14%) | 3 (15%) |
| 2 | 54 (25%) | 28 (28%) | 22 (22%) | 4 (20%) |
| 3 | 56 (25%) | 23 (23%) | 29 (29%) | 4 (20%) |
| 4+ | 52 (24%) | 18 (18%) | 28 (28%) | 6 (30%) |
| Antihypertensives taken at home | ||||
| ACEi/ARBs | 119 (54%) | 48 (48%) | 57 (57%) | 14 (70%) |
| Dihydropyridine CCBs | 85 (39%) | 35 (35%) | 43 (43%) | 7 (35%) |
| Non‐dihydropyridine CCBs | 9 (4%) | 4 (4%) | 4 (4%) | 1 (5%) |
| Thiazide diuretics | 29 (13%) | 14 (14%) | 12 (12%) | 3 (15%) |
| Beta blockers | 127 (58%) | 54 (54%) | 62 (62%) | 11 (55%) |
| Loop diuretics | 41 (19%) | 10 (10%) | 28 (28%) | 3 (15%) |
| Alpha blockers | 43 (20%) | 18 (18%) | 21 (21%) | 4 (20%) |
| Vasodilators | 77 (35%) | 33 (33%) | 37 (37%) | 7 (35%) |
| Documented medication non‐adherence | 18 (8%) | 15 (16.85%) | 3 (3%) | 1 (5%) |
| Comorbidities (via medical record) | ||||
| Hypertension | 210 (95%) | 93 (93%) | 99 (99%) | 18 (90%) |
| Previous HTN crisis hospitalizations | 22 (10%) | 4 (4.30%) | 16 (16%) | 2 (10%) |
| Type 1 or 2 diabetes | 93 (42%) | 30 (30%) | 53 (53%) | 10 (50%) |
| Hyperlipidemia | 138 (63%) | 66 (66%) | 59 (59%) | 13 (65%) |
| Current or former smoker | 88 (40%) | 41 (41%) | 42 (42%) | 5 (25%) |
| Heart failure | 55 (25%) | 11 (11%) | 36 (36%) | 8 (40%) |
| Chronic kidney disease | 81 (37%) | 21 (21%) | 54 (54%) | 6 (30%) |
| ESRD requiring dialysis | 40 (18%) | 7 (33.33%) | 30 (30%) | 3 (15%) |
| Coronary artery disease | 77 (35%) | 25 (25%) | 44 (44%) | 8 (40%) |
| Presence of cardiac devices | 14 (6.4%) | 6 (6.12%) | 7 (7%) | 1 (5%) |
| Cerebrovascular accident/transient ischemic attack | 50 (23%) | 20 (20%) | 28 (28%) | 2 (10%) |
| Triage vitals on admission | ||||
| Systolic BP (mmHg), mean (SD) | 212 (22) | 206 (21) | 217 (23) | 209 (26) |
| Diastolic BP (mmHg), mean (SD) | 101 (20) | 96 (17) | 105 (23) | 100 (14) |
| Heart rate (BPM), mean (SD) | 79 (18) | 75 (13) | 84 (21) | 77 (13) |
| Use of supplementary O2 | 13 (6%) | 2 (2%) | 8 (8%) | 3 (15%) |
| End‐organ damage signs on admission | ||||
| Serum BUN (mg/dL), median (IQR) a | 20 (15–30) | 18 (14–25) | 24 (16–41) | 19 (13–24) |
| Serum creatinine (mg/dL), median (IQR) a | 1.1 (0.9–1.8) | 1.0 (0.8–1.2) | 1.4 (1.1–5.0) | 1.0 (0.8–1.2) |
| Left ventricular hypertrophy b | 38 (17%) | 4 (26.67%) | 31 (31%) | 3 (15%) |
| Pulmonary edema b | 33 (15%) | 3 (5%) | 26 (26%) | 4 (20%) |
| Acute intracranial bleeding b | 1 (0.5%) | 0 (0%) | 1 (1%) | 0 (0%) |
| Treatment | ||||
| Admission to critical care unit | 50 (23%) | 5 (5%) | 41 (41%) | 4 (20%) |
| IVP BP‐lowering medication | 129 (59%) | 49 (49%) | 70 (70%) | 10 (50%) |
| IV drips for BP‐lowering medications | 60 (27%) | 6 (6%) | 50 (50%) | 4 (20%) |
| Discharged with new anti‐hypertensive medications | 88 (40%) | 33 (33%) | 56 (56%) | 9 (45%) |
Note: All values are retrieved from the medical records, except for demographics which were obtained from Medicare claims data.
Abbreviations: ACEi/ARB, angiotensin converting enzyme inhibitors or angiotensin receptor blocker; CCB, calcium channel blocker; ED, emergency department; ESRD, end‐stage renal disease; HTN, hypertension; ICU, intensive care unit; IQR, inter‐quartile range; IV, intravenous; IVP, intravenous push; SD, standard deviation.
N = 7 missing. Numbers presented are among those not on dialysis. It is unclear how elevated from baseline these admission creatinine are due to lack of prior laboratory data for the majority of the sample.
EOD was defined as any of the following noted in medical records: (1) neurological: new neurological deficits, altered mental status, CT/MRI radiology report of acute ischemic or hemorrhagic stroke (i.e., acute intracranial bleeding), other intracranial hemorrhage, increased intracranial pressure, head trauma; vision changes, evidence of papilledema, retinal hemorrhages, grade III/IV hypertensive retinopathy; (2) cardiac: new EKG changes (new left ventricular hypertrophy per EKG read or treating physician interpretation), structural echocardiogram changes, elevated troponins, diagnosis of acute coronary syndrome (unstable angina, non‐ST elevation myocardial infarction [NSTEMI], STEMI); (3) renal: acute kidney injury (serum Cr creatinine (Cr) increase by ≥0.3 or 50% increase in serum Cr compared to serum Cr before last non‐HTN crisis BP reading), acute renal failure, anuria/oliguria; (4) vascular: CTA/MRACT or MRI angiography evidence of aortic dissection, CXRChest X‐ray (CXR)/CT evidence of widened mediastinum, double aortic contour; (5) pulmonary: flash pulmonary edema on CXR/CT [5, 6, 9].
Among the 100 patients with an ICD‐10 claims diagnosis of HTN urgency (I16.0), the mean age was 75 years (SD, 12), and 24% were male. The mean SBP and DBP were 206 (SD, 21) and 96 (SD, 17). Approximately 93% of patients had diagnosed HTN, 91% were prescribed antihypertensive medications, 42% were prescribed 3 or more antihypertensive medications, and 4% had prior hospitalizations for HTN crisis (Table 2). The most common symptoms present on admission were headache (44%), dizziness (26%), chest pain (20%), or shortness of breath (12%); however, 29% of patients were asymptomatic. Overall, 5% eventually were admitted to a critical care unit, 49% were administered an IV push antihypertensive medication, and 6% were started on IV drip medications.
The PPV of the ICD‐10 code for HTN emergency was 91% (95% CI, 85%–96%) using the treating physician diagnosis GS, and 78% (71%–86%) using the adjudicated diagnosis GS (BP≥180/110 + evidence of EOD). Adding the ICD‐10 code for unspecified HTN crisis to HTN emergency modestly decreased the PPV to 86% (79%–92%) when using treating physician diagnosis as the GS, and 78% (71%–86%) using the adjudicated GS. The PPV was <50% to identify severe HTN emergency as determined via adjudicated diagnosis GS and critical care unit admission. The PPV for HTN urgency (I16.0) was 97% (93%–99%) using the treating physician diagnosis GS, remaining unchanged 96% (86%–99%) when restricting to ED visits only. The PPV (95% CI) for HTN urgency using an adjudicated diagnosis GS (BP≥180/110 + no evidence of EOD) was 81% (72%–88%). Similarly, the PPV was 84% (74%–91%) for HTN urgency when restricting to ED visits only. Finally, the PPV for overall HTN crisis hospitalization (either primary HTN urgency, emergency, or unspecified crisis; I16.x) was 99% (97%–100%) when using treating physician diagnosis, and 94.6% (91%–97%) using adjudicated diagnosis (Table 3).
TABLE 3.
Positive predictive value of ICD‐10 codes for hypertensive urgency (I16.0), emergency (I16.1), or unspecified crisis (I16.9).
| GS: Adjudicated diagnoses | ||||
|---|---|---|---|---|
| ICD‐10 codes / visit type | GS: Physician diagnosis a | BP + EOD b | BP + EOD + critical care c | |
| Hypertensive crisis | I16.x | 99% (97%–100%) | 94.6% (91%–97%) d | |
| Hypertensive emergency | I16.1 only | 91% (85%–96%) | 78% (70%–86%) | 48% (38%–58%) |
| I16.1 only | 91% (85%–96%) | 78% (70%–86%) | 48% (38%–58%) | |
| I16.1 or I16.9 | 86% (79%–92%) | 78% (71%–86%) | 44% (35%–53%) | |
| Hypertensive urgency | ED visits and inpatient hospitalizations | 97% (93%–99%) | 81% (72%–88%) b | N/A |
| ED visits only | 96% (89%–99%) | 84% (74%–91%) b | N/A | |
Note: Values are presented as positive predictive value (95% CI).
Abbreviations: ED, Emergency Department; EOD, End‐organ damage; GS, Gold Standard.
GS: Diagnosis reported by treating physician.
GS Adjudicated Diagnosis: BP + EOD for Hypertensive Emergency: SBP ≥180 or DBP ≥110 (thereafter, referred to BP≥180/110) + evidence of end‐organ damage; for Hypertensive Urgency, BP ≥180/110 + no evidence of end‐organ damage.
GS Adjudicated Diagnosis: BP + EOD + Critical Care: BP >180/110 + evidence of end‐organ damage + admission to ICU/CCU OR use of IV drip BP‐lowering medication.
Hypertensive crisis refers to a combined sample of both hypertensive emergency and urgency visits. The adjudicated GS for hypertensive crisis is from BP≥180/110.
When comparing concordance between the treating physician diagnosis and adjudicated diagnosis GS for HTN emergency, 78 out of 100 records (78%) had agreement. Of the 22 charts that were discordant, 17 records (55%) were deemed not to have HTN emergency per treating physician but deemed to have HTN emergency per adjudicated diagnosis whereas 45% were deemed to have HTN emergency per treating physician but did not meet adjudicated diagnosis criteria (Figure 2). Of records with descriptive reasons for the discrepancy, they included: lack of clear EOD (e.g., resolved focal neurological deficit, altered mental status prior to admission), or alternate diagnosis contributing to EOD (e.g., incidental acoustic neuroma causing headache). The overall agreement for HTN urgency was 81 out of 100 charts (81%), with 16 records (16%) deemed to have HTN urgency per treating physician but did not meet adjudicated diagnosis criteria due to BP <180/110 or evidence of EOD.
FIGURE 2.
Pairwise agreement among gold standard definitions for hypertensive urgency and emergency.
This figure shows the percent agreement and Cohen's Kappa for physician diagnosis and adjudicated diagnosis gold standards (GS). Physician diagnosis: Diagnosis reported by treating physician. Adjudicated Diagnosis for Hypertensive Emergency: SBP ≥180 or DBP ≥110 (thereafter, referred to BP≥180/110) + evidence of end‐organ damage; for Hypertensive Urgency, BP ≥180/110 + no evidence of end‐organ damage.
4. Discussion
In our national Medicare sample of 220 HTN crisis cases, the PPV was 97% for the HTN urgency (I16.0) ICD‐10 code, 91% for HTN emergency (I16.1), and 99% for overall HTN crisis (I16.x) as compared to the treating physician's diagnosis. When compared to adjudicated diagnoses as the GS, the PPV was 81% for the HTN urgency (I16.0), 78% for HTN emergency (I16.1), and 95% for overall HTN crisis (I16.x). Adding unspecified HTN crisis (I16.9) to HTN emergency (I16.1) minimally decreased the PPVs.
Prior studies on HTN crisis hospitalizations have primarily utilized expert consensus‐based ICD‐9 codes [4, 6, 7] or the specific ICD‐10 codes that have not been validated [4, 10]. No peer‐reviewed published studies have been conducted validating ICD‐10 codes; our study is the first one in a national sample of older adults demonstrating high PPV in detecting true HTN crisis, as well as HTN urgency and emergency separately. For HTN, the inclusion of specific ICD‐10 codes for HTN urgency and emergency provides a unique opportunity for targeted healthcare utilization research.
The PPVs of HTN urgency and emergency in Medicare data are similar to those in recent studies examining the validity of other cardiovascular conditions using ICD‐10 codes. These include stroke [11], heart failure [9], acute myocardial infarction [12, 13, 14, 15, 16, 17], cardiac arrests [18], arrhythmias [19], and other chronic cardiovascular conditions [20, 21, 22]. The lower PPV for HTN emergency compared to HTN urgency likely stems from the complex, often ambiguous, nature of attributing EOD to HTN [23, 24, 25, 26]. For example, a patient having an elevated admission Cr may have multiple sources of acute kidney injury, including sepsis, dehydration, progression of chronic disease. that are unrelated to HTN [27]. This usually becomes a clinical judgement that may lead to discrepancy. Lack of full clinical context for abnormal results may also explain the 5% of records with HTN emergency we observed with discordant GS—namely, no treating physician diagnosis but adjudicated diagnosis of HTN emergency.
Additionally, patient symptoms on admission such as headache, dizziness, or chest pain may be interpreted as EOD and thus given a designation of HTN emergency instead of HTN urgency. We employed strict criteria for EOD to maintain reliability across adjudicators; however, this is not likely to reflect real world practice where clinical judgement supersedes strict criteria; in fact, 17% of records we observed with a treating physician diagnosis of HTN emergency did not meet adjudicated GS. Furthermore, there may be a possibility of upcoding physician/hospital coding behaviors that could lead to increased HTN emergency ICD‐10 coding (I16.1) [28]. Finally, adding the ICD‐10 code for unspecified HTN crisis (I16.9), which may capture proportionally more ambiguous cases of EOD, to I16.1 (HTN emergency) only minimally decreased PPV for HTN emergency.
There are several limitations: (1) we studied Medicare FFS data which may not be generalizable to other populations, including privately insured, charity care, or younger populations; (2) a sizeable proportion of patients with Medicare and HTN crisis had end‐stage renal disease (ESRD), potentially limiting the generalizability to non‐ESRD patients [29]; and (3) we were only able to calculate PPV and do not know the sensitivity of these ICD‐10 codes in capturing all HTN crisis cases. Additionally, physician/hospital coding behaviors may change over time especially as new ICD‐10 codes become increasingly used which could change the PPV of the ICD‐10 codes [30].
The findings of this validation study suggest that ICD‐10 codes can be used to identify incident HTN urgency and overall HTN crisis events in outpatient ED/inpatient Medicare data with excellent PPV, and HTN emergency events in inpatient Medicare data with moderate‐to‐high PPV. HTN emergency likely requires a more nuanced approach given lack of consensus amongst providers and considerable ambiguity on the definition of EOD. Further direction in national guidelines may be necessary to standardize definitions and guide subsequent management. Investigators seeking to study incident HTN crisis risk among Medicare beneficiaries may consider using the new designated ICD‐10 codes as a validated approach to identify HTN crisis events.
Funding
Funding for this study was provided by the National Institutes of Health, National Institute on Aging (1R01AG060232‐01A1). S.S. discloses having received research funding from Pfizer Inc, Pfizer Japan, Bristol Myers Squibb, and Daiichi Sankyo and served as a consultant for Pfizer Japan and Merck Co. Inc. P.P.S. is supported by grants from National Institute of Health/National Heart, Lung and Blood Institute (Awards 1R01HL173998‐01A1, 3U01HL088942‐17S, and 1P50MD017356‐01) and National Science Foundation United States [Award (#2125872)]. He has served on the Advisory Board of RCE Technologies and HeartSciences and holds stock options; received grants or contracts from RCE Technologies, HeartSciences, Butterfly, and MindMics; and hold patents with Mayo Clinic (US8328724B2), HeartSciences (US11445918B2), and Rutgers Health (62/864,771; US202163152686P; WO2022182603A1; US202163211829P; WO2022266288A1; and US202163212228P).
Supporting information
Supporting file 1: jch70217‐sup‐0001‐tablesS1‐S2.docx
Supporting file 2: jch70217‐sup‐0002‐Abstraction‐Form.docx
Acknowledgments
The authors have nothing to report.
Contributor Information
Aayush Visaria, Email: aayush.visaria@rutgers.edu.
Soko Setoguchi, Email: ss2894@rwjms.rutgers.edu.
References
- 1. Richardson L. C., Vaughan A. S., Wright J. S., and Coronado F., “Examining the Hypertension Control Cascade in Adults With Uncontrolled Hypertension in the US,” JAMA Network Open 7, no. 9 (2024): e2431997. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Muntner P., Hardy S. T., Fine L. J., et al., “Trends in Blood Pressure Control Among US Adults With Hypertension, 1999–2000 to 2017–2018,” JAMA 324, no. 12 (2020): 1190–1200. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Chobufo M. D., Rahman E. U., Farah F., et al., “10‐Year Community Prevalence and Trends of Severe Asymptomatic Hypertension Among Patients With Hypertension in the USA: 2007–2016,” International Journal of Cardiology Hypertension 7 (2020): 100066. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Lu Y., Wang Y., Spatz E. S., et al., “National Trends and Disparities in Hospitalization for Acute Hypertension Among Medicare Beneficiaries (1999–2019),” Circulation 144, no. 21 (2021): 1683–1693. [DOI] [PubMed] [Google Scholar]
- 5. Kumar N., Simek S., Garg N., et al., “Thirty‐day Readmissions After Hospitalization for Hypertensive Emergency,” Hypertension 73, no. 1 (2019): 60–67. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Shah M., Patil S., Patel B., et al., “Trends in Hospitalization for Hypertensive Emergency, and Relationship of End‐organ Damage With in‐hospital Mortality,” American Journal of Hypertension 30, no. 7 (2017): 700–706. [DOI] [PubMed] [Google Scholar]
- 7. Janke A. T., McNaughton C. D., Brody A. M., Welch R. D., and Levy P. D., “Trends in the Incidence of Hypertensive Emergencies in US Emergency Departments From 2006 to 2013,” Journal of the American Heart Association 5, no. 12 (2016): e004511. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Hsu M. C., Wang C. C., Huang L. Y., Lin C. Y., Lin F. J., and Toh S., “Effect of ICD‐9‐CM to ICD‐10‐CM Coding System Transition on Identification of Common Conditions: An Interrupted Time Series Analysis,” Pharmacoepidemiology and Drug Safety 30, no. 12 (2021): 1653–1674. [DOI] [PubMed] [Google Scholar]
- 9. Bates B. A., Akhabue E., Nahass M. M., et al., “Validity of International Classification of Diseases (ICD)‐10 Diagnosis Codes for Identification of Acute Heart Failure Hospitalization and Heart Failure With Reduced versus Preserved Ejection Fraction in a National Medicare Sample,” Circulation: Cardiovascular Quality and Outcomes 16, no. 2 (2023): e009078. [DOI] [PubMed] [Google Scholar]
- 10. Liberman A. L., Razzak J., Lappin R. I., et al., “Risk of Major Adverse Cardiovascular Events After Emergency Department Visits for Hypertensive Urgency,” Hypertension 81, no. 7 (2024): 1592–1598. [DOI] [PubMed] [Google Scholar]
- 11. Columbo J. A., Daya N., Colantonio L. D., et al., “Derivation and Validation of ICD‐10 Codes for Identifying Incident Stroke,” JAMA Neurology 81, no. 8 (2024): 875–881. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Saunders‐Hastings P., Heong S. W., Srichaikul J., et al., “Acute Myocardial Infarction: Development and Application of an ICD‐10‐CM‐based Algorithm to a Large US Healthcare Claims‐based Database,” PLoS ONE 16, no. 7 (2021): e0253580. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Patel A. B., Quan H., Welsh R. C., et al., “Validity and Utility of ICD‐10 Administrative Health Data for Identifying ST‐and Non‐ST‐elevation Myocardial Infarction Based on Physician Chart Review,” Canadian Medical Association Open Access Journal 3, no. 4 (2015): E413–E418. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Ando T., Ooba N., Mochizuki M., et al., “Positive Predictive Value of ICD‐10 Codes for Acute Myocardial Infarction in Japan: A Validation Study at a Single Center,” BMC Health Services Research 18 (2018): 1–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Nedkoff L., Lopez D., Hung J., et al., “Validation of ICD‐10‐AM Coding for Myocardial Infarction Subtype in Hospitalisation Data,” Heart, Lung and Circulation 31, no. 6 (2022): 849–858. [DOI] [PubMed] [Google Scholar]
- 16. Tsai T. Y., Lin J. F., Tu Y. K., et al., “Validation of ICD‐10‐CM Diagnostic Codes for Identifying Patients With ST‐Elevation and Non‐ST‐Elevation Myocardial Infarction in a National Health Insurance Claims Database,” Clinical Epidemiology 15 (2023): 1027–1039. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Slezak J., Bruxvoort K. J., Sy L. S., et al., “Validation of ICD‐10 Diagnosis Codes for Identification of Acute Myocardial Infarction from a US Integrated Healthcare System,” Pharmacoepidemiology and Drug Safety 34, no. 7 (2025): e70179. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Tsai M. J., Tsai C. H., Pan R. C., Hsu C. F., and Sung S. F., “Validation of ICD‐9‐CM and ICD‐10‐CM Diagnostic Codes for Identifying Patients With out‐of‐hospital Cardiac Arrest in a National Health Insurance Claims Database,” Clinical Epidemiology 14 (2022): 721–730. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Chamberlain A. M., Roger V. L., Noseworthy P. A., et al., “Identification of Incident Atrial Fibrillation From Electronic Medical Records,” Journal of the American Heart Association 11, no. 7 (2022): e023237. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Kuang A., Xu C., Southern D. A., Sandhu N., and Quan H., “Validated Administrative Data Based ICD‐10 Algorithms for Chronic Conditions: A Systematic Review,” Journal of Epidemiology and Population Health 72, no. 4 (2024): 202744. [DOI] [PubMed] [Google Scholar]
- 21. Bose S., Stonko D. P., Kiang S. C., et al., “Validation of ICD‐10 Codes to Distinguish between Claudication and Chronic Limb‐Threatening Ischemia in Patients Undergoing Peripheral Vascular Intervention Using Medicare‐Matched Registry Data,” Circulation: Cardiovascular Quality and Outcomes 18 (2025): e011467. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Gillmeyer K. R., Nunez E. R., Rinne S. T., Qian S. X., Klings E. S., and Wiener R. S., “Development and Validation of Algorithms to Identify Pulmonary Arterial Hypertension in Administrative Data,” Chest 159, no. 5 (2021): 1986–1994. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Bress A. P., Anderson T. S., Flack J. M., et al., “The Management of Elevated Blood Pressure in the Acute Care Setting: A Scientific Statement From the American Heart Association,” Hypertension 81, no. 8 (2024): e94–106. [DOI] [PubMed] [Google Scholar]
- 24. Miller J. B., Hrabec D., Krishnamoorthy V., Kinni H., and Brook R. D., “Evaluation and Management of Hypertensive Emergency,” BMJ 386 (2024): e077205. [DOI] [PubMed] [Google Scholar]
- 25. Jones D. W., Ferdinand K. C., Taler S. J., et al., “2025 AHA/ACC/AANP/AAPA/ABC/ACCP/ACPM/AGS/AMA/ASPC/NMA/PCNA/SGIM Guideline for the Prevention, Detection, Evaluation and Management of High Blood Pressure in Adults: A Report of the American College of Cardiology/American Heart Association Joint Committee on Clinical Practice Guidelines,” Circulation 152, no. 11 (2025): e114–218. [DOI] [PubMed] [Google Scholar]
- 26. Papadopoulos D. P., Mourouzis I., Thomopoulos C., Makris T., and Papademetriou V., “Hypertension Crisis,” Blood pressure 19, no. 6 (2010): 328–336. [DOI] [PubMed] [Google Scholar]
- 27. Grams M. E., Waikar S. S., MacMahon B., Whelton S., Ballew S. H., and Coresh J., “Performance and Limitations of Administrative Data in the Identification of AKI,” Clinical Journal of the American Society of Nephrology 9, no. 4 (2014): 682–689. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Crespin D., Dworsky M., Levin J., Ruder T., and Whaley C. M., “Upcoding Linked to up to Two‐Thirds of Growth in Highest‐Intensity Hospital Discharges in 5 States, 2011–19: Study Examines Hospital Patient Discharges With Highest‐Intensity Charge Codes,” Health Affairs 43, no. 12 (2024): 1619–1627. [DOI] [PubMed] [Google Scholar]
- 29. Pothuru S., Chan W. C., Mehta H., et al., “Burden of Hypertensive Crisis in Patients With End‐Stage Kidney Disease on Maintenance Dialysis: Insights From United States Renal Data System Database,” Hypertension 80, no. 4 (2023): e59–67. [DOI] [PubMed] [Google Scholar]
- 30. Nelson S. J., Yin Y., Trujillo Rivera E. A., et al., “Are ICD Codes Reliable for Observational Studies? Assessing Coding Consistency for Data Quality,” Digital health 10 (2024): 20552076241297056. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supporting file 1: jch70217‐sup‐0001‐tablesS1‐S2.docx
Supporting file 2: jch70217‐sup‐0002‐Abstraction‐Form.docx