Heart Attack: Unmasking the Silent Killer - Rumours in Epidemiology

Be data Driven or perish

Survival Analysis
Generalized linear models
Statistical modeling
Heart Attack
Author

Bongani Ncube

Published

2 November 2024

Introduction

The study aimed to analyze survival rates in heart failure patients admitted to the Institute of Cardiology and Allied Hospital Faisalabad, Pakistan, between April and December 2015. Participants were aged 40 or above, had left ventricular systolic dysfunction, and were classified under NYHA class III and IV. The study focused on 299 heart failure patients, including 105 women and 194 men, all over the age of 40 with left ventricular systolic dysfunction, classified as NYHA class III and IV. The follow-up period ranged from 4 to 285 days, averaging 130 days. Diagnoses were based on cardiac echo reports or physician notes. Variables considered for explaining CHD mortality included age, serum sodium, serum creatinine, gender, smoking, blood pressure, ejection fraction (EF), anemia, platelets, creatinine phosphokinase (CPK), and diabetes. Continuous variables were age, serum sodium, and CPK, while categorical variables included EF, serum creatinine, and platelets. EF was categorized into three levels (EF ≤30, 30< EF ≤45, and EF >45), and platelets were divided into three levels based on quartiles.

Snapshot of our initial dataset

TIME Event Gender Smoking Diabetes BP Anaemia Age Ejection.Fraction Sodium Creatinine Pletelets CPK
97 0 0 0 0 0 1 43 50 135 1.30 237000 358
180 0 1 1 1 0 1 73 30 142 1.18 160000 231
31 1 1 1 0 1 0 70 20 134 1.83 263358 582
87 0 1 0 0 0 1 65 25 141 1.10 298000 305
113 0 1 0 0 0 0 64 60 137 1.00 242000 1610
10 1 1 0 0 0 1 75 15 137 1.20 127000 246

Explanatory data analysis

Correlation

comment
  • There are no significant correlations among these variables

Exploring Age

  • distribution of Age
#> Minimum: 40.00
#> Mean: 60.83
#> Median: 60.00
#> Mode: 60.00
#> Maximum: 95.00

comment
  • The minimum age for patients in this study is 40 while the maximum is 95.
  • On average the patients were aged 60 years

Distribution of Ejection Fraction

comment
  • The plot above shows the number of patients in each Ejection Fraction Category.
  • the majority of patients where in the 30 <EF<=45 category

Mortality Rate

comment
  • the study shows that 32.1% of the patients died and the remaining censored or were lost to followup.

Event vs Ejection Fraction

Comments
  • among the patients who died from Heart Attack , majority of the patients had Ejection Fraction less than 30 (EF<=30)

A Closer look at Ejection Fraction.

EFraction_cat n_dead n_total pct_dead low high pct
EF<=30 (93) 51 93 0.5483871 0.4701115 0.6643073 0.3110368
EF>45 (60) 14 60 0.2333333 0.1875604 0.4047309 0.2006689
30-45 (146) 31 146 0.2123288 0.1735042 0.3083021 0.4882943

Comments
  • out of the total number of people in each category , we get the proportion of patients who died.
  • it can be noted that for patients who had Ejection Fraction less than 30 , a number of them died to heart attack.

Statistical Analysis

Chi squared tests

Characteristic alive, N = 2031 dead, N = 961 Test Statistic p-value2
BP

1.9 0.17
    No 137 / 203 (67%) 57 / 96 (59%)

    Yes 66 / 203 (33%) 39 / 96 (41%)

platelets_cat

3.5 0.17
    < Q1 61 / 203 (30%) 39 / 96 (41%)

    < Q2 73 / 203 (36%) 27 / 96 (28%)

    < Q3 69 / 203 (34%) 30 / 96 (31%)

EFraction_cat

32 <0.001
    30-45 115 / 203 (57%) 31 / 96 (32%)

    EF<=30 42 / 203 (21%) 51 / 96 (53%)

    EF>45 46 / 203 (23%) 14 / 96 (15%)

Diabetes

0.00 0.97
    No 118 / 203 (58%) 56 / 96 (58%)

    Yes 85 / 203 (42%) 40 / 96 (42%)

Smoking

0.05 0.83
    No 137 / 203 (67%) 66 / 96 (69%)

    Yes 66 / 203 (33%) 30 / 96 (31%)

Anaemia

1.3 0.25
    No 120 / 203 (59%) 50 / 96 (52%)

    Yes 83 / 203 (41%) 46 / 96 (48%)

Gender

0.01 0.94
    F 71 / 203 (35%) 34 / 96 (35%)

    M 132 / 203 (65%) 62 / 96 (65%)

1 n / N (%)
2 Pearson’s Chi-squared test
comment

The chi-squared test of association tells us that only Ejection Fraction is significantly associated to the Outcome event at 5% level of significance.

take a closer look at the previous finding

Cross Tab Heart Attack x Ejection Fraction
event_cat EFraction_cat Total
30-45 EF<=30 EF>45
alive 115
56.7 % 		42
20.7 % 		46
22.7 % 		203
100 %
dead 31
32.3 % 		51
53.1 % 		14
14.6 % 		96
100 %
Total 146
48.8 % 		93
31.1 % 		60
20.1 % 		299
100 % 
#> 
#> --------Summary descriptives table by 'event_cat'---------
#> 
#> ______________________________________________________________ 
#>                   alive       dead           OR        p-value 
#>                   N=203       N=96                             
#> ¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯ 
#> EFraction_cat:                                                 
#>     30-45      115 (56.7%) 31 (32.3%)       Ref.        Ref.   
#>     EF<=30     42 (20.7%)  51 (53.1%) 4.46 [2.54;7.98] <0.001  
#>     EF>45      46 (22.7%)  14 (14.6%) 1.13 [0.54;2.30]  0.735  
#> ¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯
comment
  • The prelimenary analysis using statistical inference shows that

Patients with EF less than 30 have increased odds of dying from heart attack [2.54 ; 7.98] , this result is also statistically significant at 5% level of significance.

  • we will find more on the models that come.

Boxplots

comment
  • it should be noted that there is some difference in Ejection Fraction between between people who died and those who survived.
  • we will try to answer more using a t.test

T test

Characteristic alive, N = 2031 dead, N = 961 Difference2 95% CI2,3 p-value2
Age 59 (11) 65 (13) -6.5 -9.5, -3.4 <0.001
Sodium 137.2 (4.0) 135.4 (5.0) 1.8 0.69, 3.0 0.002
Creatinine 1.18 (0.65) 1.84 (1.47) -0.65 -0.96, -0.34 <0.001
Ejection.Fraction 40 (11) 33 (13) 6.8 3.9, 9.7 <0.001
TIME 158 (68) 71 (62) 87 72, 103 <0.001
Pletelets 266,657 (97,531) 256,381 (98,526) 10,276 -13,723, 34,276 0.4
CPK 540 (754) 670 (1,317) -130 -416, 156 0.4
1 Mean (SD)
2 Welch Two Sample t-test
3 CI = Confidence Interval
Note

Comments

  • The table above shows that there is significant difference in mean Age,Sodium,Creatinine and mean Ejection Fraction implying that the differences in the averages is not merely an outcome of chance

Taking the variables one at a time

comment
  • The plot above shows the categories of patients who are at the risk of experiencing heart attack
  • Patients who have an Ejection Fraction less than 30 (*EF<=30*) have a higher risk of experiencing or dying from heart attack.
  • Patients aged 70 and above (*70_inf*) also face a higher risk of dying from Heart attack
  • Patients whose platelets are below the lower quartile (*<_Q1*) and have BP and anaemia have higher chances of dying from heart attack.
  • much is not clear about effect of diabetes , but it turns out not to be a major cause of death.

Statistical Modeling

Generalised linear models (Logistic Regression)

Univariate Model

Characteristic N OR1 95% CI1 p-value
platelets_cat 299


    < Q1
— —
    < Q2
0.58 0.32, 1.05 0.072
    < Q3
0.68 0.38, 1.22 0.2
Gender 299 0.98 0.59, 1.64 >0.9
Age 299 1.05 1.03, 1.07 <0.001
EFraction_cat 299


    30-45
— —
    EF<=30
4.50 2.57, 8.04 <0.001
    EF>45
1.13 0.54, 2.28 0.7
Smoking 299 0.94 0.56, 1.58 0.8
Sodium 299 0.91 0.85, 0.96 0.001
BP 299 1.42 0.86, 2.34 0.2
Diabetes 299 0.99 0.60, 1.62 >0.9
Anaemia 299 1.33 0.82, 2.17 0.3
CPK 299 1.00 1.00, 1.00 0.3
1 OR = Odds Ratio, CI = Confidence Interval

Multivariate Model

Characteristic OR1 95% CI1 p-value
platelets_cat


    < Q1 — —
    < Q2 0.44 0.21, 0.90 0.027
    < Q3 0.66 0.33, 1.31 0.2
Gender 0.68 0.34, 1.35 0.3
Age 1.07 1.04, 1.10 <0.001
EFraction_cat


    30-45 — —
    EF<=30 5.86 3.10, 11.5 <0.001
    EF>45 1.06 0.47, 2.32 0.9
Smoking 1.02 0.52, 2.00 >0.9
Sodium 0.91 0.85, 0.97 0.004
BP 1.41 0.79, 2.53 0.2
Diabetes 1.14 0.64, 2.06 0.7
Anaemia 1.23 0.68, 2.23 0.5
CPK 1.00 1.00, 1.00 0.013
No. Obs. 299

Log-likelihood -150

AIC 325

BIC 373

1 OR = Odds Ratio, CI = Confidence Interval
comment

  • For a unit increase in Age , the odds of dying increase by a significant percentage > Patients with Ejection Fraction below 30 have 6 times chances of dying from Heart attack as comprared to other groups.

  • A unit increase in sodium , decreases the odds of dying from Heart Attack

comment
  • The plot above shows variables that are most important in prediction of heart attack. Ejection Fraction , Age and Sodium are the top 3 most important predictors of heart attack.
#> # Indices of model performance
#> 
#> AIC     |    AICc |     BIC | Tjur's R2 |  RMSE | Sigma | Log_loss | Score_log | Score_spherical |   PCP
#> --------------------------------------------------------------------------------------------------------
#> 325.141 | 326.419 | 373.247 |     0.238 | 0.408 | 1.000 |    0.500 |   -43.829 |           0.012 | 0.668

#> Call: survfit(formula = Surv(TIME, Event) ~ 1, data = out_new)
#> 
#>  time n.risk n.event survival std.err lower 95% CI upper 95% CI
#>     4    299       1    0.997 0.00334        0.990        1.000
#>     6    298       1    0.993 0.00471        0.984        1.000
#>     7    297       2    0.987 0.00664        0.974        1.000
#>     8    295       2    0.980 0.00811        0.964        0.996
#>    10    293       6    0.960 0.01135        0.938        0.982
#>    11    287       2    0.953 0.01222        0.930        0.977
#>    13    284       1    0.950 0.01263        0.925        0.975
#>    14    283       2    0.943 0.01340        0.917        0.970
#>    15    281       2    0.936 0.01412        0.909        0.964
#>    20    278       2    0.930 0.01480        0.901        0.959
#>    23    275       2    0.923 0.01545        0.893        0.954
#>    24    273       1    0.920 0.01575        0.889        0.951
#>    26    272       3    0.909 0.01663        0.877        0.943
#>    27    269       1    0.906 0.01691        0.873        0.940
#>    28    268       2    0.899 0.01745        0.866        0.934
#>    29    266       1    0.896 0.01771        0.862        0.931
#>    30    264       4    0.882 0.01869        0.846        0.920
#>    31    259       1    0.879 0.01893        0.843        0.917
#>    32    258       1    0.875 0.01916        0.839        0.914
#>    33    257       2    0.869 0.01961        0.831        0.908
#>    35    254       1    0.865 0.01983        0.827        0.905
#>    38    253       1    0.862 0.02004        0.823        0.902
#>    40    252       1    0.858 0.02025        0.820        0.899
#>    41    251       1    0.855 0.02046        0.816        0.896
#>    42    250       1    0.852 0.02066        0.812        0.893
#>    43    249       3    0.841 0.02124        0.801        0.884
#>    44    246       1    0.838 0.02143        0.797        0.881
#>    45    245       1    0.834 0.02161        0.793        0.878
#>    50    244       1    0.831 0.02179        0.789        0.875
#>    55    241       1    0.828 0.02197        0.786        0.872
#>    59    240       1    0.824 0.02215        0.782        0.869
#>    60    239       2    0.817 0.02250        0.774        0.863
#>    61    236       1    0.814 0.02267        0.771        0.859
#>    64    234       1    0.810 0.02283        0.767        0.856
#>    65    233       2    0.803 0.02316        0.759        0.850
#>    66    231       1    0.800 0.02332        0.755        0.847
#>    67    230       1    0.796 0.02348        0.752        0.844
#>    72    227       1    0.793 0.02364        0.748        0.841
#>    73    225       2    0.786 0.02394        0.740        0.834
#>    77    217       1    0.782 0.02411        0.736        0.831
#>    78    216       1    0.779 0.02427        0.732        0.828
#>    82    207       1    0.775 0.02444        0.728        0.824
#>    88    194       1    0.771 0.02464        0.724        0.821
#>    90    189       2    0.763 0.02504        0.715        0.813
#>    95    180       1    0.758 0.02526        0.711        0.810
#>    96    175       1    0.754 0.02548        0.706        0.806
#>   100    173       1    0.750 0.02571        0.701        0.802
#>   109    159       1    0.745 0.02597        0.696        0.798
#>   111    155       1    0.740 0.02625        0.691        0.794
#>   113    152       1    0.735 0.02652        0.685        0.789
#>   115    150       1    0.730 0.02679        0.680        0.785
#>   126    136       1    0.725 0.02713        0.674        0.780
#>   129    135       1    0.720 0.02746        0.668        0.776
#>   130    134       1    0.714 0.02777        0.662        0.771
#>   135    132       1    0.709 0.02808        0.656        0.766
#>   150    118       1    0.703 0.02848        0.649        0.761
#>   154    117       1    0.697 0.02886        0.643        0.756
#>   162    116       1    0.691 0.02923        0.636        0.751
#>   170    115       1    0.685 0.02959        0.629        0.745
#>   171    114       1    0.679 0.02993        0.623        0.740
#>   172    113       2    0.667 0.03059        0.610        0.730
#>   180    106       2    0.654 0.03128        0.596        0.719
#>   193     86       1    0.647 0.03183        0.587        0.712
#>   196     83       1    0.639 0.03238        0.578        0.706
#>   198     79       1    0.631 0.03297        0.569        0.699
#>   207     71       1    0.622 0.03368        0.559        0.692
#>   214     53       1    0.610 0.03503        0.545        0.683
#>   235     37       1    0.594 0.03776        0.524        0.673
#>   241     33       1    0.576 0.04068        0.501        0.661
comment
  • looking at the survival curve ,note that it always starts at 1 and goes down in time

Summary by Gender

Note
  • there are minor differences in the survival times (time to death) between males and females

Summary by Ejection Fraction

comment
  • Patients with Ejection Fraction less than 30 experience death faster as compared to the other 2 groups.
  • it can be noted from the log rank that the survival times between these groups is significantly different

Summary by Platelets

comment
  • we notice a small difference in these curves starting
  • The difference is not statistically different

Cox proportional Hazard

Characteristic HR1 95% CI1 p-value
platelets_cat

0.2
    < Q1 — —
    < Q2 0.64 0.38, 1.08
    < Q3 0.78 0.47, 1.29
Gender 0.77 0.48, 1.26 0.3
Age 1.05 1.03, 1.07 <0.001
EFraction_cat

<0.001
    30-45 — —
    EF<=30 3.46 2.16, 5.55
    EF>45 1.22 0.64, 2.33
Smoking 1.09 0.67, 1.77 0.7
Sodium 0.93 0.89, 0.97 0.002
BP 1.73 1.13, 2.66 0.013
Diabetes 1.08 0.70, 1.65 0.7
Anaemia 1.33 0.86, 2.06 0.2
CPK 1.00 1.00, 1.00 0.018
No. Obs. 299

Log-likelihood -470

AIC 964

BIC 995

1 HR = Hazard Ratio, CI = Confidence Interval
Comments
  • holding all other covariates constant ,increase in Age increases the hazard of death by 5%
  • an decrease in Ejection Fraction increases the hazard of death by 4 folds , we see this because the hazard of death is four fold for patients with Ejection Fraction less than 30.

Visualise the results

Deductions
  • Age and Ejection Fraction are significant predictors of death
  • the higher Age groups have a higher probability of dying from heart attack
  • The statistical analysis identified age, EF, creatinine, sodium, anemia and BP as the significant variables affecting the likelihood of mortality among heart failure patients.
Conclusion

It can be concluded that growing age, renal dysfunction (having serum creatinine greater than its normal level 1.5), high BP (higher than normal range), higher level of anaemia and lower values of ejection fraction (EF) are the key factors contributing towards increased risk of mortality among heart failure patients. Increased level of serum sodium can reduce the odds of death. No significant differences were found due to smoking status, diabetes and gender of patients.