Delta HR/BT as a possible predictor of severity or mortality, not a predictor of bacterial infection

Dear Editor,

We read with interest the article by Hamano and Tokuda [1] on changes in vital signs as predictors of bacterial infection. The authors concluded the delta heart rate (HR)/body temperature (BT) could help physicians determine whether home care patients have bacterial infections. However, the study’s conclusion is discordant with the result. They presented the sensitivity, specificity, positive predictive value, negative predictive value, and the overall accuracy of delta HR/BT with various cutoff points, on the diagnosis of bacterial infections. We are reminded of two famous mnemonics: ‘SnNout,’ which means, ‘When a diagnostic test has an extremely high sensitivity, a negative test result rules out the target disorder,’ and ‘SpPin,’ which means, ‘When a diagnostic test has an extremely high specificity, a positive test result rules in the target disorder [2].’ However, since sensitivity and specificity are a trade-off relationship, judgement may be wrong if interpreted separately. When we calculate the likelihood ratios (LRs) from the sensitivity and specificity they presented, the positive LRs and negative LRs are around 0.7–1.5, as shown in Table 1. LRs between 0.5 and 2 make little changes in the probability of disease [3]. Predictive values depend on the prevalence on the target disorder; even poor diagnostic tests have a high positive predictive value when the disorder is common [3]. For example, given the prevalence of bacterial infection, which is 70.6% in patients with fever in the study, positive predictive values around 70% mean that the test has little influence on the probability.

Table 1. Sensitivity, specificity, positive, and negative LRs of delta HR/BT.

	Cutoff	Sensitivity (%)	Specificity (%)	Positive LR	Negative LR
Patients with fever
Averaged delta HR/BT	≥5	75.9	28.1	1.06	0.86
	≥10	54.7	49.1	1.07	0.92
	≥15	31.4	68.4	0.99	1.00
	≥20	20.4	84.2	1.29	0.95
	≥25	13.9	89.5	1.32	0.96
Assumed delta HR/BT	≥5	91.2	8.9	1.00	0.99
	≥10	77.2	17.9	0.94	1.27
	≥15	55.1	42.9	0.96	1.05
	≥20	36.8	62.5	0.98	1.01
	≥25	22.1	78.6	1.03	0.99
Patients with fever <38℃
Averaged delta HR/BT	≥5	68.1	27.3	0.94	1.17
	≥10	52.8	50	1.06	0.94
	≥15	34.7	65.9	1.02	0.99
	≥20	23.6	79.5	1.15	0.96
	≥25	19.4	86.4	1.43	0.93
Assumed delta HR/BT	≥5	93	9.3	1.03	0.75
	≥10	83.1	14	0.97	1.21
	≥15	66.2	37.2	1.05	0.91
	≥20	47.9	58.1	1.14	0.90
	≥25	33.8	74.4	1.32	0.89

HR: heart rate; BT: body temperature; LR: likelihood ratio.

The study about the relationship between HR and BT dates back to the beginning of the twentieth century. Mackenzie noted, ‘Roughly speaking there is an increase of ten beats with each rise in temperature of one degree Fahrenheit [4],’ which corresponded to an increase of about 20 bpm accompanied by a rise in temperature of 1℃. In 1927, Lyon confirmed this relationship between HR and BT not only in patients with bacterial infections but also in patients with a non-infectious process such as rheumatoid arthritis and epilepsy [5]. Therefore, the pulse–temperature relationship is considered a physiologic response to inflammation rather than a specific response to bacterial infection.

Lyon also noticed that this correlation between HR and BT was low in fatal cases of pneumonia [5]. Leibovici et al. reported that the pulse–BT ratio was an independent predictor of 30-day mortality [6]. Taken together, delta HR/BT could be a possible predictor of severity or mortality, not a predictor of diagnosis of bacterial infection. Further study to examine the association between delta HR/BT and severity or mortality is warranted.

Declaration of interest

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Additional information

Funding

This paper was not funded.