HAVOC: new score developed to identify AF patients at higher risk of stroke


The HAVOC score was developed following the CRYSTAL-AF study, to see if there was an easier way to identify which patients with atrial fibrillation were most at risk of stroke. Susan Zhao, Santa Clara Valley Medical Center, San Jose, USA, presented an evaluation of the HAVOC score at the International Stroke Conference (ISC; 24–26 January 2018, Los Angeles, USA).

Atrial fibrillation (AF) is associated with ischaemic stroke, heart failure and death, and the identification of AF in the post-stroke population is very important in reducing future AF-related morbidity and mortality, particularly in patients with cryptogenic stroke. A high incidence of AF has been identified using insertable cardiac monitors (ICMs) in patients with recent strokes of unknown causes, mainly from CRYSTAL-AF study.

Further identifying subsets of patients who will benefit from ICMs is desirable and urgently needed. This led Zhao and her colleagues to develop the HAVOC score.

“To give you some background on how we came up with the HAVOC score, we used the Stanford STRIKE database which is a fully index-able, searchable, validated database including 20 million patient encounters and we isolated close to 10,000 patients over the age of 40 with cryptogenic stroke or transient ischaemic stroke. Using coding we identified close to 500 patients within this cohort who developed AF post-stroke diagnosis and with COX multivariable digression analysis we developed risk factors: hypertension, age, valvular heart disease, vascular disease, obesity, congestive heart failure, coronary artery disease,” said Zhao. “Out of those seven predictors we constructed a score called HAVOC and it is weighted on the strength of their correlation coefficient, regression coefficient. So patients with a low HAVOC score (0–4) vs. medium (5–9) vs. high (10–14) have a significant rate of differences in detection of AF. With a low score at only 2.5% over an average of three years follow up whereas a high score, those with a score of 10–14 will have a close to 25% chance of AF over a close the three year follow-up.”

The HAVOC score was constructed from a retrospective cohort study. To validate it the group applied it to the CRYSTAL-AF study data. They looked to see if the HAVOC score could also predict AF among cryptogenic stoke patients in the CRYSTAL-AF study who were continuously monitored by ICM.

Out of the 441 patients that were randomised, there were 221 in the ICM arm and 220 in the control arm and a total of 214 patients were analysed. Once the HAVOC score system was applied to the 214 suitable patients it turned out that most of the 214 patients fell into the low risk HAVOC score category. To gain further insight the 214 patents were further divided into three groups: group A, group B and group C. Group A patients scored about 0–1 on the HAVOC score, and group B scored 2–3 and group C scored ≥4. There was 66 in the low group, 104 in the middle group and 44 in the high group.

In group A obesity was the only factor that contributed to their HAVOC score and 22.7% were obese compared to 34.1% in the overall group.

When the HAVOC score was applied to the 214 patients in the CRYSTAL-AF study, the mean score was 2.4, the median score was 2 and 89% had a score of ≤4. The mean HAVOC score tended to be higher among patients with AF—2.9±1.8 vs. 2.3±2.1 (p=0.07)—but this was not statistically significant.

In group A patients the changes of an AF diagnosis were very low, only 11% of them were diagnosed with AF after 12 months of monitoring. In group B (104 patients), 18% of them were diagnosed with AF after 12 months, whereas in the high score group (44 patients) 32% of patients went on to be diagnosed with AF and this reached statistical significance with p=0.02.

Of the individual components of the HAVOC score (hypertension, age, valvular heart disease, vascular disease, obesity, congestive heart failure, coronary heart disease), age was the strongest predictor of AF development in the stroke population.

Overall the conclusion for this analysis was that the vast majority of patients in the CRYSTAL-AF study had low HAVOC scores (≤4).

“This potentially explains the differences between AF detection rates between CRYSTAL-AF and other studies of cryptogenic stroke patients,” Concluded Zhao. “HAVOC scores should be considered when comparing AF detection rates between different studies.”

Mean HAVOC scores tended to be higher in patients with AF vs. those without AF, however it didn’t quite reach significance but this could be due to the very small sample size. AF incidence increased significantly with increasing HAVOC scores meaning that the HAVOC score could be a useful approach to help guide cardiologists and neurologists to identify patients who could potentially benefit more from prolonged rhythm more from prolonged rhythm monitoring studies and tests.


Please enter your comment!
Please enter your name here