Participants for the HMHS were recruited from the NHANES sample – a cross-sectional survey of the non-institutionalized USA civilian population. NHANES collects extensive health information through home interviews and physical measurements. Further details on the NHANES sample design and data collection methods can be found elsewhere [4].

HMHS participants were a convenience sample recruited during a 4-month period in the NHANES 2011–2012 survey cycle. Participants were enrolled only after they completed the following examinations required at the MEC: anthropometry, BP, and phlebotomy. Interested participants signed an informed consent electronically. Home visits were scheduled in the first available time slot after the MEC visit (normally 1–3 weeks later) as close to the same time of day as the MEC exam as possible. However, flexibility was allowed in scheduling home exams if needed to avoid refusals. Pregnant women and nonfluent English speakers were excluded from participation in the study. A total of 130 participants aged 18 years and older participated in the HMHS. Further details of the HMHS study are provided elsewhere [5]. BP was not measured for two participants because of equipment malfunction in the home, resulting in a final analytic sample of 128 participants.

The National Center for Health Statistics Research Ethics Review Board reviewed and approved the HMHS study design and all participants provided written consent.

In this study, home BP measurement refers to a single assessment at a particular point in time obtained in the home by a trained examiner, rather than self-monitoring when measurements are obtained (by the participant him/herself) multiple times within a given day for several days.

The FI and the HT performed the home exam on the same visit. HTs had, at minimum, an Associate’s degree in Health Science, cardiopulmonary resuscitation certification, and at least 1-year experience working in the healthcare field. FIs had, at minimum, a high school diploma or equivalent but no previous healthcare experience or training.

In the home, data collection included dried blood spots, height, weight, and BP. The approximate duration to complete one exam sequence was 30 min. Details of the study design and descriptions of all three components can be found elsewhere [5]. A brief description of the BP component is presented here.

Participants were not told which examiner was the FI or the HT. Similarly, home examiners were instructed not to volunteer their skill level. Examiners were not allowed to observe each other perform the exam sequence; thus, the home examiners were not aware of each other’s BP readings. The second examiner started his or her exam only after the first examiner completed his or her exam and packed up his or her equipment. In addition, the order of the examiners was randomized to control for any order effect [6].

Two weeks before the start of the HMHS, four FIs and two HTs were trained on the NHANES BP protocol [7] using the same device. The BP measurements were three successive readings on the same arm, with the participants seated with their backs supported, both feet flat on the floor, and forearms supported on a stable surface at the heart level [8]. An initial 5-min wait period was observed before the start of each BP exam, with 30-second rest intervals between each of the BP readings. The BP results were manually keyed in by the examiners as the BP devices do not have data-transfer capability. The examiners keyed in their BP results twice to reduce data entry error. All of the participants had six replicate BP readings in the home (three by the HT and three by the FI).

A subset of participants (n = 82) had an extra set of BP measurements obtained over a sleeved arm by the FI. The rationale for recording BP over the sleeve is explained in the series report [5]. BP differences between bare arm and sleeved arm are beyond the scope of this study and are not explored in this paper.

The OMRON HEM-907XL Intellisense digital BP monitor (Omron Healthcare, Bannockburn, Illinois, USA) was used to obtain BP. The Omron device was chosen for the study because it had been validated in a previous NHANES methodology study [9]. It has a ‘hide’ feature that was used during the home BP measurements to block visibility of the measurements from the participant. Cuff size in the home was determined by estimating the mid-arm circumference (mid-AC) using a sex-specific regression equation published by Ostchega et al. [10] to minimize physical contact with a participant. This equation used the height and weight measurements obtained in the anthropometry section of the home exam to calculate mid-AC. Mid-AC was then used to select the appropriate cuff size. The Omron machine had four available cuff sizes fitting the following arm circumferences: small (17.0–22.0 cm), adult (>22.0–32.0 cm), large adult (>32.0–42.0 cm), and extra-large adult (>42.0–50.0 cm).

The FI performed monthly calibrations on all the Omron machines using the Netech Digimano 2000 pressure vacuum (Netech Corp., Farmingdale, New York, USA). The calibration technique involved connecting the Omron device and the pressure vacuum gauge through a T-connector [8,9]. Pressure points were compared between the two devices and comparative readings were required to fall within ± 3 mmHg [8,9]. Before the start of the HMHS, all BP machines passed calibration testing. This responsibility for calibration was intentionally given to the least experienced examiner to determine whether equipment maintenance could be performed by FIs.

As the HMHS participants were drawn from the NHANES sample, all HMHS participants had BP measurements taken in the MEC as part of their routine physical examination by a trained physician using the auscultation method using a standard wall-mounted mercury sphygmomanometer and a Littman Cardiology III stethoscope (3M Health Care, St. Paul, Minnesota, USA). The physician followed the standard NHANES protocol described on the NHANES website [7] to obtain three replicate BP measurements. The appropriate cuff size was based on a direct measurement of mid-AC [11]. There were four available cuff sizes for the mercury sphygmomanometer fitting the following arm circumferences: small (17–21.9 cm), adult (22–29.9 cm), large adult (30–37.9 cm), and extra-large (38–47.9 cm). The physician keyed in all BP values by hand. BP equipment in the MEC are inspected on a regular basis.

In the first set of analyses, the examiners in the home were compared. Mean differences were calculated and compared with national standards. Differences in BP classification were also compared. The second set of analyses compared results from the FI with the MEC physician. For each examiner, averages were calculated for the three sets of replicate systolic blood pressure (SBP) and diastolic blood pressure (DBP) measures. These averages were used throughout the subsequent analyses. Overall mean BP values were computed and compared by the home examiner. The magnitude of the BP difference was assessed between the home examiners, overall, by sex, and by the age of the participants. Inter-rater reliability between the FI and HT was assessed by Pearson’s correlation coefficients.

To provide another familiar metric for the evaluation of differences between the two examiners, the mean and absolute differences by home examiner were compared with criteria for accuracy developed by the Association for the Advancement of Medical Instrumentation (AAMI) and the revised British Hypertension Society (BHS). It should be noted that the validation approach used to compare the examiners in the home are traditionally used to evaluate devices rather than individuals. The AAMI guidelines indicate that the mean difference between values measured by the two different types of examiners should be no greater than ± 5 mmHg, with a maximum SD of 8 mmHg [12,13]. The revised BHS guidelines use a grading system from A to D to evaluate accuracy from best to worse. We calculated the percentage of absolute differences between the examiners that fell within 5, 10, and 15 mmHg [14]. Bland–Altman graphs and scatter plots were used to visually show agreement between the home examiners.

We classified individuals into the following categories: elevated systolic blood pressure (ESBP), elevated diastolic blood pressure (EDBP), and HBP. ESBP was defined as SBP of 140 mmHg or more, EDBP was defined as DBP of 90 mmHg or more, and HBP was defined as both ESBP and EDBP. To assess the level of agreement of HBP classification between the two home examiners, κ-statistics and percent agreement were calculated.

In a second set of analyses, we examined the classification agreement of HBP between the FI and the MEC physician by also using the κ-statistic and percent agreement.

All tests were evaluated for statistical significance using a P value of less than 0.05. All analyses were carried out using the SAS software (version 9.3; SAS Institute, Cary, North Carolina, USA). No adjustment was made for multiple comparisons.

The overall BP mean ± SD measured by HTs (SBP, 117.0 ± 12.7 mmHg; DBP, 69.9 ± 9.2 mmHg) was lower than the overall BP mean measured by FIs (SPB, 119.0 ± 14.4 mmHg; DBP, 71.9 ± 9.8 mmHg) (Table 2).

The mean difference was 2.0 ± 6.2 mmHg (P < 0.001) for SBP and 2.0 ± 4.4 mmHg (P < 0.001) for DBP (Table 3). The FI had higher BP readings compared with the HT for both sexes and for one age group (18–39 years of age) (Table 3).

The mean BP measurements between the home examiners were highly correlated (r = 0.903 for SBP and r = 0.894 for DBP) (Fig. 1).

The Bland–Altman graphs provide a visual comparison of the average differences in SBP and DBP measures (HT vs. FI) against the corresponding average BP measurement of the examiners (Fig. 2). All figures showed some extreme values beyond two SD. The Bland–Altman plots show good agreement for both SBP and DBP with no clear linear pattern, indicating, on average, that the difference in BP readings between the HT and FI was close to the zero line along the entire continuum of BP values.

Also, in Table 3, the comparison between the FI and HT measurements fulfilled the AAMI criteria, with the exception of the SBP measurements in the 60 years and older age group (n = 39), in which the SD for the difference between the HT and FI measurements was8.5 mmHg.

The percentages of BP differences within 5, 10, and 15 mmHg were 64, 91, and 97%, respectively, for SBP and 71, 97, and 99%, respectively, for DBP (Table 3).

Table 4 presents the percentages of participants who were classified as having elevated SBP (ESBP ≥ 140 mmHg), elevated DBP (EDBP ≥ 90 mmHg), and HBP (SBP/DBP ≥ 140/90 mmHg) by examiner type. In the home, the FI classified 8.6% of participants as having ESBP, whereas the HT classified 6.3%. The FI identified 5.5% as having EDBP, whereas the HT identified 0.8%. The FI classified 11.7% of participants as having HBP, whereas the HT classified 7.0% of the participants as having HBP. The FI consistently identified more cases than the HT in all categories.

The percent agreement between the HT and FI in the determination of ESBP was 95%, with a κ-statistics of 0.60 [95% confidence interval (CI) = 0.30–0.87], suggesting moderate agreement. Similarly, the percent agreement for HBP was 92%, with a moderate κ-statistics [κ = 0.54, 95% CI = 0.30–0.79]. However, the percent agreement for EDBP was 95%, with a lower κ-statistics [κ = 0.24, 95% CI = − 0.15 to 0.63], indicating fair agreement. A κ-statistics value between 0.21 and 0.40 represents fair agreement, a value between 0.41 and 0.60 represents moderate agreement, and a value between0.61 and 0.80 represents considerable agreement [15] (Table 5).

κ-Statistic are usually used to assess observer agreement, but can be constrained by low prevalence.

The mean SBP (119.6 ± 15.4 mmHg) and the mean DBP (72.7 ± 9.3 mmHg) values from the MEC were higher than those reported in the home, irrespective of the home examiner.

In the MEC, the physician classified 10.2% of participants as having ESBP, 3.1% as having EDBP, and 10.2% as having HBP (Table 4), whereas the FI classified 8.6% of participants as having ESBP, 5.5% as having EDBP, and 11.7% of participants as having HBP.

The percent agreement across two categories (ESBP and HBP) between the FI and physician was 92 and 91%, respectively, with the following moderate κ-statistics (κ = 0.54, 95% CI = 0.29–0.79 and 0.57, 95% CI = 0.35–0.80, respectively). The percent agreement for EDBP between the physician and FI was 91%, but the κ-statistic was low (κ = 0.34, 95% CI = − 0.03 to 0.71) (Table 5).

Overall, the average BP measurements obtained by FIs with no healthcare experience were close to those obtained by HTs. Although FIs recorded higher measured BP values than the HTs, the results were nevertheless comparable with each other on the basis of the AAMI and BHS criteria. However, the HT and the FI in the home classified different percentages of HBP. Similar results in the classification of HBP were observed between the FI and the MEC physician. Currently, there is a lack of objectively measured data on cardiovascular disease at the subnational level. The inclusion of BP measurements as part of a nationally representative population survey such as NHIS could help to provide these estimates. At this time, it is difficult to tease out the most important contributing factors for the differences between the home examiners, and with a small sample of hypertensive adults, our comparisons are limited.