IHP 525 Module 6 Diagnostic Tests Paper Example

How Accurate Is Home Monitoring? Evaluating a Blood Pressure Device Against Ambulatory Readings

[Student Name]

Southern New Hampshire University

IHP 525: Biostatistics

Module Six Paper

[Instructor Name]

[Date]

How Accurate Is Home Monitoring? Evaluating a Blood Pressure Device Against Ambulatory Readings

A screening tool is only as useful as its ability to sort people correctly. Before Bayview County expands home blood pressure monitoring, program leaders need to know how often a home result agrees with the reference standard and what a positive or negative result means for the person holding it. This paper evaluates the device, explains the measures in plain terms and recommends how results should be used.

The Study Design

Four hundred program participants measured their pressure twice each morning and evening for seven days on the program's validated upper-arm device. Within two weeks, each also wore a 24-hour ambulatory monitor, which serves as the reference standard because it captures pressure through daily activity and sleep. A home average of 135 mm Hg systolic or higher counted as a positive result; a 24-hour ambulatory average of 130 or higher counted as true uncontrolled hypertension. Of the 400, 120 had uncontrolled hypertension by the reference, a prevalence of 30%.

Sensitivity and Specificity

Altman and Bland (1994a) define sensitivity as the proportion of people with the condition whom the test correctly identifies and specificity as the proportion without the condition whom it correctly clears. Home monitoring was positive in 96 of the 120 people with uncontrolled hypertension, a sensitivity of 80%. It was negative in 252 of the 280 without it, a specificity of 90%. They stress that these proportions are estimates from a sample and should carry confidence intervals.

Table 1. Home Monitoring Against 24-Hour Ambulatory Reference

Note. Composite data for 400 program participants.

How Precise Are These Estimates?

To get a standard error for any proportion, multiply p by one minus p, divide by the number of people and take the square root. For sensitivity, that is the square root of 0.8 times 0.2 over 120, or 0.0365, so the 95% interval is 80% plus or minus 7.2 points, from 72.8% to 87.2%. For specificity, the standard error is the square root of 0.9 times 0.1 over 280, or 0.0179, giving an interval from 86.5% to 93.5%. Sensitivity is less precise because only 120 people had the condition; a future study wanting a narrower interval would need more people with uncontrolled pressure, not simply more participants overall.

What a Result Means for the Person Holding It

A participant does not know whether they have uncontrolled hypertension; they only know their home result. Altman and Bland (1994b) explain that predictive values answer the question patients actually ask. Of the 124 people flagged by the device, 96 truly had uncontrolled pressure, a positive predictive value of 96 of 124, or 77%. Of the 276 whose home average stayed below 135, 252 were truly controlled, a negative predictive value of 252 of 276, or 91%. In the program, then, roughly three in four positive home results reflect true uncontrolled pressure, and about nine in ten negative results are reassuring.

Why Prevalence Changes the Picture

The same note makes a point that program leaders often miss: predictive values shift with the share of the tested group who actually have uncontrolled pressure, while sensitivity and specificity, in principle, do not. Program participants were recruited because of high readings, so prevalence was 30%. If the county offered home monitoring to all adults, where perhaps 10% have undetected uncontrolled pressure, the arithmetic changes sharply. In 1,000 adults, 100 would have the condition and 80 would test positive; of the 900 without it, 10% or 90 would also test positive. The positive predictive value would be 80 of 170, or about 47%, so more than half of positive results would be false alarms. The negative predictive value would rise to 810 of 830, about 98%.

Table 2. Predictive Values at Two Levels of Prevalence

Note. Sensitivity of 80% and specificity of 90% are assumed in both settings.

Likelihood Ratios

Likelihood ratios offer a way to carry accuracy across settings. Dividing 0.80, the sensitivity, by 0.10, the false positive rate, gives a positive likelihood ratio of 8, meaning a positive home result is eight times as likely in someone with uncontrolled pressure as in someone without it. For a negative home result, the miss rate of 0.20 over the specificity of 0.90 yields about 0.22. Clinicians can combine these with their own sense of a patient's prior probability, which is more flexible than quoting a single predictive value.

Choosing the Cutoff

The 135 threshold is a choice, not a fact of nature. Lowering it catches more true cases but also flags more people who are controlled. Altman and Bland (1994c) describe the receiver operating characteristic plot, which graphs sensitivity against one minus specificity across possible cutoffs; the closer the curve bends toward the upper left corner, the better the test discriminates, and the area under it summarizes overall accuracy. The five cutoffs examined here trace a curve with an estimated area of about 0.91, indicating strong discrimination.

Table 3. Sensitivity and Specificity Across Home Cutoffs

Note. Values estimated from the 400 composite participants.

Why 135 Remains the Right Threshold

The best cutoff depends on the costs of each kind of error. Missing uncontrolled pressure delays treatment and raises stroke risk, which argues for a lower threshold. False positives, however, lead to extra clinic visits and possible overtreatment, and program nurses have limited appointment slots. At 130, sensitivity rises to 89% but specificity falls to 80%, doubling false positives from 28 to 56 in this sample. Because a positive home result triggers a nurse review rather than an immediate medication change, and because negative results will be rechecked in three months, 135 offers a reasonable balance. It also matches the threshold recommended for home readings in current US guidance, which keeps program messages consistent with what patients hear from their doctors.

Limitations

Three limitations apply. First, all 400 were program participants, so accuracy in the general population may differ if people there have milder pressure elevations that are harder to classify. Second, home readings depend on correct technique; participants were trained, and untrained users may do worse. Third, ambulatory monitoring is itself imperfect, so some apparent errors of the home device may be errors of the reference.

Recommendations

Program staff should treat a positive home result as a reason for nurse review within two weeks, not as a diagnosis. Negative results should be repeated at three months for anyone with a previous high clinic reading. If the county considers a countywide offer, it should expect roughly half of positive results to be false and plan confirmation capacity accordingly. Training in measurement technique should accompany every device issued.

Conclusion

Home monitoring identified four in five people with uncontrolled pressure and correctly cleared nine in ten without it. Those figures are strong enough to support use within the program, but predictive values show the same device would produce many false alarms in a lower-risk population. Accuracy figures only become useful once they are read against the setting in which a test is used.

References

Altman, D. G., & Bland, J. M. (1994a). Statistics notes: Diagnostic tests 1: Sensitivity and specificity. BMJ, 308(6943), 1552. https://doi.org/10.1136/bmj.308.6943.1552

Altman, D. G., & Bland, J. M. (1994b). Statistics notes: Diagnostic tests 2: Predictive values. BMJ, 309(6947), 102. https://doi.org/10.1136/bmj.309.6947.102

Altman, D. G., & Bland, J. M. (1994c). Statistics notes: Diagnostic tests 3: Receiver operating characteristic plots. BMJ, 309(6948), 188. https://doi.org/10.1136/bmj.309.6948.188