survival-data-handler 2024.4.10

No project description provided

Survival data handler

The aim of this package is to facilitate the use of survival data by switching from temporal data in the form of a collection of survival functions to temporal matrices calculating other functions derived from survival analysis, such as residual life, hazard function, etc. analysis, such as residual life expectancy, hazard function, etc.

import pandas as pd

from lifelines import CoxPHFitter
from lifelines.datasets import load_rossi

rossi = load_rossi()
cph = CoxPHFitter()

cph.fit(rossi, duration_col='week', event_col='arrest')
curves = cph.predict_survival_function(rossi).T
curves.columns = pd.to_timedelta(curves.columns.to_numpy() * 7, unit="D")
print(curves.head())

	7 days 00:00:00	14 days 00:00:00	21 days 00:00:00	28 days 00:00:00	35 days 00:00:00	42 days 00:00:00	49 days 00:00:00	56 days 00:00:00	63 days 00:00:00	70 days 00:00:00
0	0.997616	0.99523	0.992848	0.990468	0.988085	0.985699	0.983305	0.971402	0.966614	0.964223
1	0.993695	0.987411	0.981162	0.974941	0.968739	0.962552	0.95637	0.926001	0.913958	0.907978
2	0.994083	0.988183	0.982314	0.976468	0.970639	0.96482	0.959004	0.930402	0.919043	0.913399
3	0.999045	0.998089	0.997133	0.996176	0.995216	0.994254	0.993287	0.98846	0.986508	0.985531
4	0.997626	0.99525	0.992878	0.990507	0.988135	0.985758	0.983374	0.97152	0.966752	0.96437

from survival_data_handler import Lifespan


age = pd.to_timedelta(rossi["age"] * 365.25, unit="D")
birth = pd.to_datetime('2000')
rossi["index"] = rossi.index
    
birth = pd.to_datetime('2000')

lifespan = Lifespan(
    curves,
    index=rossi["index"],
    birth=birth,
    age=age,
    window=(pd.to_datetime("2000"), pd.to_datetime("2001"))
)

We now add the supervision data (in the form of duration)

lifespan.add_supervision(
    event=rossi["arrest"],                                      # True if the data is observed False, when censored
    durations=rossi["duration"] + birth      # The duration
)

Let's calculate the associated performance

lifespan.assess_metric("survival_function")

Date	Performance (1 - AUC)
2000-01-31	0.468458
2000-03-01	0.384425
2000-03-31	0.432012
2000-04-30	0.357338
2000-05-30	0.365263
2000-06-29	0.365190
2000-07-29	0.371438
2000-08-28	0.343447
2000-09-27	0.340607
2000-10-27	0.344628
2000-11-26	0.334398
2000-12-26	0.334444